> ## Agent Instructions
> This documentation is available in English and Simplified Chinese. English pages use root paths; Chinese counterparts use the same path under /zh/.
> For a Chinese-language query, retrieve the corresponding /zh/ page when available. Preserve its data, account, permission, and execution limitations when answering.
# Driven Agents: Investment workspaces
Source: https://docs.driven.ai/concepts/agents
An Agent is a dedicated investment workspace in Driven with its own memory, research, scheduled tasks, and portfolio. Learn when to use one Agent versus many.
An **Agent** is a dedicated investment workspace that organizes conversations, research files, preferences, paper portfolios, and scheduled tasks. Select the intended Agent when continuing research or configuring its work.
## What an Agent contains
Each Agent holds:
* **Conversations and session history** — the running record of what you have asked and discussed
* **A Profile** — the Agent's identity, style, and how it works for you
* **Portfolios** — the portfolios this Agent runs, each with its own positions, cash, orders, and performance
* **Playbooks** — the investment strategy behind each portfolio
* **Generated research files** — reports and briefs the Agent has produced
* **Scheduled tasks** — recurring workflows the Agent runs on its own
* **Channel connections** — web, mobile, Telegram, WeChat
## When to create multiple Agents
Create separate Agents when two jobs should not share assumptions. The test is simple: if combining them into one workspace would muddy the context or make a scheduled task ambiguous, split them; if they naturally share the same universe and rules, keep them together.
For concrete starting patterns and how to set one up, see [Quickstart](/get-started/quickstart).
## Good Agent instructions
An Agent's standing instructions should describe stable preferences, the things that are true across every conversation:
```text theme={null}
Focus on evidence from filings, financial statements, and earnings calls. Separate facts from assumptions. Always include risks and what would invalidate the thesis.
```
Avoid instructions that are temporary or too broad. A one-off request like "analyze AAPL today" belongs in chat, not in the Agent's permanent instructions. Anything you would want applied to your hundredth conversation belongs here; anything you want once belongs in the chat.
**Memory** stores facts, research notes, and references. **Profile** holds Agent-wide preferences. **Playbook** holds one portfolio's strategy rules. These instructions are not enforced risk controls; check proposed trades.
## Profile, portfolios, and Playbooks
These three fit together. The Agent's [Profile](/get-started/quickstart) sets the Agent's identity and style, and applies across everything it does. Each [portfolio](/concepts/portfolio-and-paper-trading) the Agent runs carries its own [Playbook](/concepts/playbook) — the strategy that drives that portfolio's universe, risk rules, and trading. The Profile is Agent-wide; a Playbook is per portfolio.
## Related
* [Playbook](/concepts/playbook) — persistent strategy memory
* [Multi-agent strategies](/guides/strategy/multi-agent-strategies) — running different strategies in parallel
* [Quickstart](/get-started/quickstart) — set one up
# Driven channels: Telegram and WeChat
Source: https://docs.driven.ai/concepts/channels
Connect a Driven Agent to Telegram or WeChat, so you can chat, run Skills, switch models, and receive scheduled briefings in Telegram.
Channels let you use a Driven Agent from the messaging apps you already use. Connect an Agent to Telegram or WeChat and you can chat with it and run Skills. Telegram can also receive scheduled briefings and alerts without opening the web app.
The web app stays the home base for setup and deep review. Channels are for using your Agent on the go, in a place you already check every day.
## Supported channels
* **Telegram** — the most complete channel: rich formatting, group chats, a command menu, and scheduled-task delivery.
* **WeChat** — chat with your Agent one-on-one from WeChat.
Driven currently supports Telegram and WeChat.
## Connect a channel
A channel connects to an existing Agent, so create an Agent first. Then open the channel options from the session sidebar and follow the steps for your channel.
### Telegram
1. In Telegram, open **@BotFather**, send `/newbot`, and follow the prompts to create a bot.
2. Copy the **bot token** BotFather gives you, paste it into Driven, and click **Connect**.
3. Click **Open Telegram** to activate the bot and claim it as yours. You can now message your Agent directly in Telegram.
### WeChat
1. Choose to connect WeChat, and Driven shows a **QR code**.
2. Scan it from the WeChat app to sign in. You may be asked for a verification code.
3. Once confirmed, your Agent is reachable from WeChat.
One Agent can be connected to more than one channel. A given bot or WeChat login is tied to a single Agent at a time.
## What you can do in a channel
* **Chat** with your Agent, including sending photos, voice messages, and file attachments where supported by the channel
* **Run Skills** and ask follow-up questions
* **Receive scheduled briefings** and alerts in Telegram, with any generated files attached
* **Switch the model** your Agent uses from an owner one-on-one chat
* **Manage the conversation** with a set of commands
Channels reward short, pointed questions. Save long, multi-part research for the web app, where the output is easier to review.
For web app upload format details, see [File uploads](/reference/file-uploads).
## Channel commands
Type a command, or use Telegram's command menu, to control the conversation:
| Command | What it does |
| - | - |
| `/new` | Start a new conversation |
| `/model` | Switch the model your Agent uses |
| `/status` | Show the current conversation's status |
| `/usage` | Show your account usage and plan |
| `/compact` | Compact the conversation context |
| `/stop` | Stop the Agent's current run |
| `/skills` | List available Skills |
| `/help` | List commands |
Owner and account commands like `/model`, `/status`, and `/usage` work in a one-on-one chat with your Agent.
## Telegram and WeChat compared
Both let you chat with your Agent. They differ in what else they support:
| | Telegram | WeChat |
| - | - | - |
| One-on-one chat | Yes | Yes |
| Group chats | Yes | — |
| Rich formatting (code, tables) | Yes | Plain text |
| Command menu | Yes | Type commands as text |
| Scheduled-task delivery | Yes | Not supported |
| Send file attachments and charts | Yes | Yes |
If you rely on scheduled briefings delivered to your phone, use Telegram. WeChat is best for one-on-one questions and answers.
## Use the web app for deep review
For long research, generated files, portfolio details, scheduled-task setup, and connecting channels, the web app is easier to inspect.
Telegram and WeChat do not show web approval dialogs. Connector tools may run automatically even when web permissions require approval. Review [connector permissions](/concepts/connectors#choose-a-permission-mode) before enabling external actions.
## Related
* [Scheduled tasks](/concepts/scheduled-tasks) — set up briefings delivered to the web app or Telegram
* [Agents](/concepts/agents) — a channel connects to an Agent
* [Skills](/concepts/skills) — what you can run from a channel
# Driven connectors: Data and brokerage tools
Source: https://docs.driven.ai/concepts/connectors
Connect external data and brokerage tools to a Driven Agent. Understand web approval modes and automatic execution in scheduled or messaging workflows.
Connectors give an Agent access to tools from external services. A connector can add brokerage features, research data, productivity tools, or other capabilities provided by the service.
Each connector belongs to one Agent. Connect it separately if another Agent should use the same service.
## Connect from Explore
1. Open the Agent that should use the connector.
2. Click **Connectors**, then open **Explore**.
3. Select a connector to review its author, server URL, description, and available tools.
4. Click **Connect**.
5. Complete the service's authorization flow if prompted.
After the connection succeeds, it appears under **My connectors**.
## Add a custom connector
Use a custom connector when a service gives you a compatible server URL that is not listed in **Explore**.
1. Open **Connectors** and click **Add connector**.
2. Enter a name and the **Server URL** supplied by the connector developer.
3. If the developer requires your own OAuth application, open **Advanced settings** and enter its client ID and optional client secret. Otherwise, leave these fields empty.
4. Click **Add** and complete authorization if Driven opens an authorization page.
Only connect servers from developers you trust. Connector developers control the tools their servers expose and may change those tools over time. Review the connector URL, requested access, and available tools before using it.
## Choose a permission mode
Open a connected connector's detail page to change **Permissions**. New connectors default to **Allow read-only actions**. The approval behavior below applies to conversations in the web app.
| Mode | Behavior |
| - | - |
| **Allow read-only actions** | Driven can run tools classified as read-only. Other actions require your approval. |
| **Allow all actions** | Driven can run the connector's tools without asking each time. |
| **Ask every time** | Driven asks for approval before every connector tool call. |
Read-only classification uses the connector server's labels and tool-name patterns. Treat it as a safeguard, not a guarantee. Review approval requests carefully, especially when a tool can trade, send messages, modify records, or disclose data.
Scheduled tasks, Telegram, and WeChat do not support the web approval prompt. Connector tools can run automatically in these contexts even when the permission mode would ask for approval on the web. Disable a connector if you do not want its tools available to these workflows. Do not rely on **Ask every time** to pause a scheduled or messaging-channel action.
## Use a connector in chat
You can use a connector in four ways:
* Describe what you need in natural language. The Agent can automatically choose a relevant connected connector for the task.
* Click the **Connectors** icon in the chat composer and choose a connected service.
* Type `/` and choose a connector from the **Connectors** section.
* Open a connected connector's detail page and click **Try now**.
The selected connector appears as a chip in the composer. Write your request and send it normally. The Agent can then use that connector's tools for the request. If an action requires approval, review the proposed action and choose whether to approve or deny it.
## Manage a connector
Open **My connectors** to manage every connector attached to the current Agent.
* Turn the switch off to pause a connector without removing its stored connection.
* Turn it on to make it available again.
* Select **Re-authorize** if its authorization has expired or failed.
* Open the detail page to review the current server URL, tools, and permission mode.
* Select **Disconnect** from the `...` menu to remove the connector and its stored credentials immediately.
Connector settings do not have an edit flow. To change the server URL or OAuth client credentials, disconnect the connector and add it again.
## Troubleshooting
* **Connectors is missing or disabled:** the feature may not be enabled for your account yet.
* **Authorization did not finish:** open **My connectors** and select **Authorize** or **Re-authorize**.
* **No tools appear:** open the connector detail page. The server may currently expose no tools, or the connection may need to be retried.
* **A connector stopped working:** verify that the server URL is still current, re-authorize if prompted, and confirm the external service is available.
## Related
* [Agents](/concepts/agents) - connectors are configured separately for each Agent
* [Skills](/concepts/skills) - reusable workflows that can use available tools
* [Data sources](/reference/data-sources) - the built-in financial data available without a connector
# Driven financial data and market coverage
Source: https://docs.driven.ai/concepts/data-and-coverage
Use financial data for US, Hong Kong, and A-share research, with clear source, coverage, and timestamp boundaries.
Driven connects your Agent to financial data for research and analysis. Built-in coverage focuses on US, Hong Kong, and A-share equities, with ETF, options, macro, FX, crypto, and commodity context. The product describes 260+ data endpoints across 30+ categories; availability varies by market, field, and source.
## Main categories
| Category | Examples |
| - | - |
| Prices and instruments | Current and historical quotes, charts, ETF data, and supported US/HK options chains |
| Company research | Financial statements, valuation metrics, analyst expectations, and earnings calls |
| Ownership and activity | Institutional holdings, insider disclosures, congressional trades, and market-specific flow data |
| Events and context | News, filings, calendars, economic releases, rates, and cross-asset data |
Hong Kong and A-share statements can use local or cumulative reporting periods. Check accounting scope and currency before comparing figures. A-share coverage also includes Dragon & Tiger disclosures, margin trading, shareholder records, share pledges, and valuation bands. See [Data sources](/reference/data-sources) for categories.
## Read the time and evidence correctly
A fresh query can return an older observation. Quote time, reporting period, filing date, and retrieval time have different meanings. Illiquid contracts may retain old quotes, and institutional disclosures can lag positions.
Ask for the actual source, date, unit, currency, and missing fields. Separate retrieved facts from calculations and model judgments. Neither a data connection nor a recent timestamp guarantees a correct answer.
```text theme={null}
Analyze NVDA. Cite the statement fields, filings, and quotes used.
Show their dates, distinguish calculations from assumptions, and flag gaps.
```
## Research and trading coverage
Research availability does not establish trading support. Built-in paper trading supports US and Hong Kong accounts. Live positions and orders require supported brokerage [Connectors](/concepts/connectors), account permissions, and the relevant instruments. [Options](/concepts/options) explains chain and screening scope.
Use connectors to extend available services where supported. Do not assume an external subscription or a named data source is automatically connected to your Agent.
## Related
* [Supported markets](/reference/supported-markets)
* [Data sources](/reference/data-sources)
* [Insider Tracker](/skills/smart-money)
# Driven Options: Research and position analysis
Source: https://docs.driven.ai/concepts/options
Research US and Hong Kong options, compare strategies, sync supported brokerage positions, and review trades with clear data and permission limits.
**Options** brings contract research, strategy tools, real positions, and earlier conversations into one workspace. Continue from opportunity discovery to position analysis, tracking, and review. Live trading is available where broker tools and account permissions support it.
Research does not require a broker connection. Real positions and trade history require a supported broker. The **Options** entry and new-chat suggestions may not be enabled for every account yet.
## Open the workspace
Open your Agent and click **Options** in the sidebar. Choose positions, an analysis card, a strategy, or a conversation under **Option track**.
## Use data and compare strategies
Driven provides supported US and Hong Kong chains, expirations, and contract details. Compare candidates using strikes, bid/ask, volume, open interest, IV, and Greeks.
When data supports it, ask the Agent to calculate spreads, breakevens, expiration payoffs, and ROC. Request the capital basis, annualization, quantities, multipliers, and currencies. Expiration payoffs do not establish early-exit results, and midpoint prices do not guarantee fills.
**Strategies** provides downside-protection and income-enhancement entry points: protective puts, bear put spreads, protective collars, cash-secured puts, covered calls, and covered strangles. Clicking fills a suggested question. Add holdings, horizon, and risk constraints before sending. Other structures can be researched in chat.
```text theme={null}
Compare a protective put and protective collar for 100 AAPL shares over
one month. Show each leg, quote time, premium, breakevens, capital requirements,
and loss bounds. Label assumptions and missing data. Do not place orders.
```
Real-time access, fields, and trading sessions depend on the source. Check each quote's timestamp. Strategy illustrations and animated charts on the product page are demonstrations.
## Connect and sync positions
1. Connect and authorize a supported broker from the current Agent's **Connectors**. When no broker is connected, you can also click **Connect** in Options.
2. Return to **Options** and select the broker under **Connector**.
3. Click **Sync Positions** for the first sync, inspect the conversation and result, then return to the workspace.
4. Switch between accounts when needed. The refresh icon beside the update time starts a new conversation that refreshes the selected account.
Sync saves real options accounts, excluding paper accounts and futures-only accounts. The broker must expose the required account and position tools.
| Position column | What you see |
| - | - |
| **Options**, **Qty** | Underlying, expiration, strike, Call / Put, and quantity; short quantities are negative |
| **Cost / Bid-Ask** | Average cost and available bid/ask; a position price appears when quotes are unavailable |
| **vs Strike** | Moneyness and distance from strike when data is available |
| **Delta**, **Theta**, **IV** | Available sensitivity and implied-volatility values for each leg |
Positions are saved snapshots. Update time is not each contract's quote time. Refresh before using them. Missing metrics are not estimated; the table is not a complete account valuation, margin balance, or performance history.
## Analyze positions and trades
| Card | Use it for |
| - | - |
| **Leg-by-Leg Analysis** | Each leg's risk, Greeks, payoff, and strategy contribution |
| **Options Portfolio Analysis** | Holdings and combined risk in the selected account |
| **Options Trade Review** | Decisions using the last month's order and fill history |
The first two request a position refresh. If it fails, the Agent can use an old snapshot and should flag its age. Trade review needs history; current positions cannot establish fills.
## Research contracts and opportunities
General contract screening currently covers US options. For Hong Kong research, start with a named underlying's chain. Describe filters directly or use a specialized Skill:
| Skill / entry | Scope and limits |
| - | - |
| **Unusual Option Activities** | US contract snapshots; they cannot establish trade direction, institutional intent, or multi-leg execution |
| **Earnings Option Opportunities** / **Earnings Option Analysis** | Earnings timing, event coverage, liquidity, and volatility; defaults to US, permits an explicit Hong Kong request, and covers one market per scan |
| **0DTE Option Opportunities** / **Scan 0DTE Options** | Contracts expiring on the current US trading day, requiring verified expirations and usable quotes |
```text theme={null}
Research US earnings options over the next two weeks. Check event timing,
expirations, and liquidity before comparing strategies. Separate facts,
calculations, and judgments. Explain when no opportunity qualifies.
```
Zero qualifying opportunities is valid. Outside regular hours, treat 0DTE results as structural reference and revalidate after the open. Protection ends at that day's expiration.
## Track and review weekly
**Option track** lists the current Agent's options conversations. Reopen one to continue research, refresh quotes, and test earlier assumptions.
Click **Weekly Wrap**, inspect the suggested request, add a preferred time or destination if needed, then send it. The default is Saturday at 09:00 in your timezone, covering the preceding seven days. Check the Agent's confirmed schedule, timezone, and next run.
The review uses web options conversations, saved positions, and available broker fills. It covers position changes, executed trades, discussed opportunities, and your stated plans. It does not sync positions automatically or audit every channel. Refresh first if you need recent positions; old snapshots and missing history should appear in data notes. Manage it through [Scheduled tasks](/concepts/scheduled-tasks).
## Live trading and risk
Research, strategy cards, and sync do not place orders. Live trading requires options order tools, account authorization, and appropriate permissions. Check the account, each leg, opening/closing intent, quantities, limits, multipliers, currencies, and funding requirements. After submission, inspect fill status.
Web approval follows [connector permissions](/concepts/connectors#choose-a-permission-mode). Scheduled tasks, Telegram, and WeChat do not support web approval prompts; tools can run automatically. Do not rely on **Ask every time** to pause orders in those contexts. State execution boundaries and disable the broker connector when needed.
Options can lose their entire premium, and some short strategies can incur unlimited losses. Verify liquidity, margin, assignment risk, and expiration. Analysis is informational. Investment decisions and responsibility remain yours.
## Related
* [Driven Options product introduction](https://driven.ai/options)
* [Connectors](/concepts/connectors)
* [Skills](/concepts/skills)
* [Options risk introduction (SEC)](https://www.investor.gov/introduction-investing/general-resources/news-alerts/alerts-bulletins/investor-bulletins-63)
# Driven scheduled tasks and market monitoring
Source: https://docs.driven.ai/concepts/scheduled-tasks
Schedule Driven research and watchlist checks. Set time zones and delivery channels, and understand usage, notification, and connector permissions.
Scheduled tasks let an Agent run prompts automatically on a schedule. Instead of opening Driven and asking the same question every morning, you define the question once and the Agent runs it on its own, delivering the result to your chosen channel.
Tasks can run while the app is closed, subject to account access, available usage, and service availability. They check data at scheduled times rather than monitoring every market tick.
Conditions in a prompt guide analysis; they are not notification switches. Scheduled conversation results may still be sent to a configured channel, including a no-events summary. Test runs and delivery; change task delivery settings if you do not want Telegram pushes.
## What you can automate
* Daily pre-market briefing
* Watchlist scan for overnight moves and news
* Earnings summary on report days
* Portfolio risk check
* Unusual volume monitor
* Macro calendar summary
## Example
```text theme={null}
Every weekday at 8 AM New York time, scan my watchlist and send a pre-market briefing. Include major news, price moves, earnings events, and risks to watch today.
```
## What a good scheduled task includes
The more precise the task, the more useful the output. A complete task usually specifies:
* **Frequency** — daily, weekly, every weekday, etc.
* **Time zone** — anchor the time explicitly (e.g. "New York time")
* **Universe or watchlist** — which names or markets to cover
* **Output format** — bullets, a short brief, a ranked list
* **Delivery channel** — web or Telegram
* **Stop conditions or alert thresholds** — only notify if something crosses a line
Specify thresholds, comparison windows, and how to handle missing data. Test runs and delivery to confirm conditional notification behavior.
## Delivery channels
Scheduled task output can be delivered to the web app or to Telegram. For a morning briefing you read on your phone, route it to Telegram. For heavier output you want to inspect closely, the web app is easier.
Scheduled tasks do not show web approval dialogs. Connector tools may run automatically even when web permissions require approval. Define execution limits and disable connectors the task should not use. See [connector permissions](/concepts/connectors#choose-a-permission-mode).
## Review the logs
Review task runs periodically. If the output is too broad, tighten the prompt or narrow the universe. If a task stops being useful, edit or retire it. Treat scheduled tasks like a living part of your process, not set-and-forget infrastructure.
## Related
* [Create a market briefing](/guides/scheduled/create-a-market-briefing) — full walkthrough
* [Set up watchlist monitoring](/guides/scheduled/set-up-watchlist-monitoring) — threshold-based alerts
* [Advanced scheduled tasks](/guides/scheduled/advanced-scheduled-tasks) — conditional and multi-step tasks
* [Scheduled task prompts](/prompts/scheduled-task-prompts) — copyable templates
# Driven Skill Store: Install and share Skills
Source: https://docs.driven.ai/concepts/skill-store
Publish, review, install, and update reusable Skills in Driven.
Skill Store lets creators share reusable Skills and lets other users install approved versions into their own Agents. A Skill you build in an Agent workspace starts as a local draft. After you publish it and it passes review, it becomes a Store version that other users can install.
## How it works
Each time you publish or update a Skill, Driven creates a separate version and reviews it. Approved versions appear in **Store** and can be installed into an Agent. If a newer version is approved later, users with an older installed version see an update notice and choose when to update.
In short:
* A draft Skill belongs to the Agent where you create it.
* A Store version must pass review before other users can install it.
* An installed Skill becomes available only to the selected Agent.
* Updates require user confirmation.
## For creators
Build and test the Skill in the Agent where you created it. Before publishing, make sure `SKILL.md` clearly explains what the Skill does, when to use it, and what kind of result to expect.
When the Skill is ready, select **Publish**. If the Skill already has a Store version and the local files changed, select **Publish update** from **My skills** or **Publish changes** from **Published**. Driven creates a new version and adds it to review.
Driven reviews every submission after you publish. Use the **Published** tab to track review status, view version history, and publish later changes.
During review:
* The submitted version does not appear in **Store** yet.
* The current published version stays available while an update is under review.
* A rejected update does not replace the current published version.
* Only one version of the same Skill can be pending review at a time.
## Review
Review checks whether a Skill is ready for Store. Driven checks whether the Skill is complete, whether its behavior matches its description, and whether it raises safety or privacy concerns.
If a version is rejected, open the review details and inspect the issues and evidence. Select **Edit in chat** to ask Driven to help update the current Skill files, then test the Skill again before publishing another version.
## For users
Open **Store** to browse approved Skills. A Skill can also be opened from a shared Skill link. When you find a Skill you want to use, select **Install** and choose the Agent that should use it.
Installed Skills are managed from **My skills**. From there, you can:
* enable or disable a Skill
* see whether a newer version is available
* select **Update** to sync the latest approved version
* uninstall a Skill you no longer need
Installation applies only to the Agent you choose. Install the same Skill separately if another Agent should use it too.
## Related
* [Skills](/concepts/skills) - what Skills are and how to invoke them
* [Skills reference](/skills/overview) - built-in Skills available in Driven
* [Skill routing logic](/reference/skill-routing-logic) - how Driven chooses a Skill for your question
# Driven Skills: Reusable investing workflows
Source: https://docs.driven.ai/concepts/skills
Use reusable investing workflows with defined methods, data requirements, and output checks.
A **Skill** gives your Agent instructions for a specific research workflow. It can define a method, data requirements, steps, output format, and checks. Some Skills include calculation scripts; the Agent still needs usable data and must follow the instructions correctly.
## Use a Skill
Describe your task in chat, choose an enabled Skill from the composer, or type `/` to find an available Skill. You can also name it in your request:
```text theme={null}
Use Valuation Matrix to value MSFT. Show assumptions, sources, and missing data.
```
The Agent chooses workflows based on the task and available Skills. A named Skill can hand off to another workflow when the request is outside its scope. Simple lookups may use data tools directly. See [Skill routing](/reference/skill-routing-logic).
## Choose a workflow
* Company research: **Stock Analysis** or **Deep Research**
* Valuation: **Valuation Matrix**
* Same-industry competitive ranking: **Competitor Analysis**
* Stock discovery: **Stock Screener**
* Macro context: **Market Pulse**
* US sector and industry rotation: **Sector Radar**
* Disclosed ownership activity: **Insider Tracker**
* Portfolio risk: **Portfolio Monitor**
* Options research: [dedicated options Skills](/concepts/options#research-contracts-and-opportunities)
See the [Skills reference](/skills/overview) for the current catalog and scope.
## Create and install Skills
Create a Skill to reuse your own process, or install an approved community Skill from **Skill Store** into the Agent that should use it. Review its purpose and requirements, and enable it before use. See [Skill Store](/concepts/skill-store).
## Check the result
Ask for sources, calculation assumptions, and missing data. A defined workflow helps you review the work; it does not guarantee identical answers or eliminate model errors. You remain responsible for investment decisions.
# Driven plans and pricing
Source: https://docs.driven.ai/get-started/plans-and-pricing
Compare Free, Pro, Max, and Ultra monthly pricing, workspace limits, and model access.
Driven offers Free, Pro, Max, and Ultra plans. The table shows published monthly offers. See [current pricing](https://driven.ai/pricing) for annual offers and current quotes.
| | Free | Pro | Max | Ultra |
| - | - | - | - | - |
| Monthly price | Free | \$20/mo | \$100/mo | \$200/mo |
| Displayed usage | Limited | Standard | 5x | 20x |
| Models | Auto | Selected models | More advanced models | More flagship models |
| Agents / user | 1 | 1 | 3 | 6 |
| Paper portfolios / Agent | 1 | 3 | 5 | 10 |
| Storage / Agent | 1 GB | 2 GB | 4 GB | 8 GB |
| Scheduled tasks | Within plan usage | Within plan usage | Within plan usage | Within plan usage |
| Early access to new features | — | — | Yes | Yes |
| Priority access at peak times | — | — | — | Yes |
Prices and base limits were checked on **October 9, 2026**. Confirm the current offer and checkout details in your account and on the pricing page.
## Choose a plan
* **Free**: start research with Auto and limited usage.
* **Pro**: everyday research with more usage and model choice.
* **Max**: deeper research or several separate Agents.
* **Ultra**: higher usage, more workspaces, and flagship-model access.
## Usage conditions
Agents, data, Skills, memory, and scheduled tasks form the core workflow. Specific features depend on rollout, data sources, and account permissions. External brokerage or data services can require separate authorization or fees; a Driven subscription does not include every external entitlement.
Usage multipliers are plan comparisons, not guaranteed question counts. Models, tasks, and tools affect consumption. Model access also depends on region and current versions. See [Limits and quotas](/reference/limits-and-quotas).
## Billing and support
Check the offer in your account and the final checkout details before purchase. For upgrades, downgrades, cancellation, or billing, use in-app support or the [pricing FAQ](https://driven.ai/pricing).
# Get started with Driven
Source: https://docs.driven.ai/get-started/quickstart
Sign in with Google or an email code, open a Driven Agent, run your first investment research task, and save Profile preferences.
Start with one Agent and one concrete research question, then save your language and research preferences.
## 1. Sign in
Open Driven and choose either sign-in method:
* **Google**: Click **Sign in with Google** and complete authorization.
* **Email verification code**: Enter your email address, click **Continue with email**, then enter the six-digit code sent to your inbox and click **Sign in**. No password is required.
You can start with the Free plan. If the code does not arrive or has expired, see [sign-in troubleshooting](/reference/troubleshooting#sign-in-problems).
## 2. Create or open an Agent
Use an existing **Agent**, or choose the new-Agent entry from the Agent menu. The flow creates your workspace automatically; you do not need configuration files. Start with one Agent, then add others for different preferences or separate research contexts. Agent limits depend on your [plan](/get-started/plans-and-pricing).
## 3. Ask a concrete question
The quality of your answer tracks the specificity of your question. Start with something concrete:
```text theme={null}
Analyze AAPL after the latest earnings report. Focus on revenue growth, margin trend, valuation, and risks.
```
For broader research, point the Agent at a theme:
```text theme={null}
Research listed companies in the AI semiconductor supply chain. Explain comparison criteria, show source dates, and separate facts, assumptions, and risks. Do not place orders.
```
The Agent can use enabled Skills or data tools to work on the question. State your desired output and ask it to explain the evidence and workflow used.
## 4. Review the answer
For investment work, the answer is only as good as its evidence. Check:
* **Sources** — are the financial figures pulled from data, not guessed?
* **Assumptions** — does the answer separate facts from interpretation?
* **Risks** — are the downside cases concrete and specific?
* **What would change the view** — does it say what evidence would flip the conclusion?
If anything is thin, push back in the same conversation: ask for the bear case, ask it to date its claims, or ask it to cite the specific statement fields it used.
## 5. Tell the Agent how you invest
Give the Agent a name and save its purpose, language, and research preferences in chat:
```text theme={null}
Name this Agent Sage. Save to its Profile: focus on US and Hong Kong
technology companies, use concise English, prioritize financial statements,
and separate facts, assumptions, and risks. I invest with a multi-year horizon.
```
Ask it to read the saved Profile back. Agent-wide preferences belong in **Profile**. The rules for one portfolio belong in that portfolio's [Playbook](/concepts/playbook); saved research facts belong in memory. Preferences do not enforce risk controls, so inspect later results.
## Next steps
* [Run deep research](/guides/research/run-deep-research) — multi-source research with exportable reports
* [Create a market briefing](/guides/scheduled/create-a-market-briefing) — your first scheduled task
* [Plans and pricing](/get-started/plans-and-pricing) — compare Pro, Max, and Ultra
# What is Driven? AI investment Agents
Source: https://docs.driven.ai/get-started/what-is-driven
Driven connects investment research, options analysis, market monitoring, portfolio strategies, and supported brokerage trading in an AI Agent workspace.
Driven is an AI Agent platform for investors. Use conversation to research markets, analyze options and positions, build strategies, schedule monitoring, and execute trades through supported brokerage connections.
**Your investment team, driven by you.** The Agent gathers data, calculates, analyzes, and runs tasks. You set goals, inspect evidence, and decide whether to act.
## Your investment workspace
An **Agent** contains conversations, research files, memory, and scheduled tasks. Set its research scope and communication preferences, or create separate Agents for distinct goals. Agent limits and model access depend on your [plan](/get-started/plans-and-pricing) and region.
**Profile** holds Agent-wide language, style, and stable preferences. **Memory** holds facts and research notes. **Playbook** holds the strategy rules for a specific portfolio. See [Agents](/concepts/agents).
Driven includes Claude, GPT, Gemini, Grok, and other model families so you can start research without assembling the model and data tools yourself. Use [DrivenBench](/why-driven/drivenbench) to compare capability, cost, and latency when choosing a model.
## Core capabilities
| Capability | What you can do |
| - | - |
| [Professional data and research](/concepts/data-and-coverage) | Use 260+ financial data endpoints for US, Hong Kong, and A-share research, with multi-asset and macro context |
| [Options analysis](/concepts/options) | Query US/HK chains, screen contracts, compare strategies, analyze real positions, and track research |
| [Professional Skills](/concepts/skills) | Reuse valuation, screening, research, and risk methods; install community Skills through [Skill Store](/concepts/skill-store) |
| [Scheduled monitoring](/concepts/scheduled-tasks) | Run briefings, watchlist checks, and risk reviews on a schedule, even after closing the conversation |
| [Strategy management](/guides/strategy/build-your-playbook) | Save a portfolio's universe, sizing, and risk rules, then revisit them with new evidence |
| [Broker connections](/concepts/connectors) | Read supported real accounts and positions, then use live order tools where permissions allow |
Coverage, refresh rates, and real-time entitlements vary by source. Monitoring follows the task schedule rather than observing every market tick. Live trading requires broker, account, and regional support. Built-in paper trading uses virtual funds.
## Start with one question
```text theme={null}
Research NVDA's latest earnings and valuation. Separate facts from assumptions.
Compare growth, margins, and key risks. Include sources and data timestamps.
```
Inspect the result, then add your horizon and risk preferences. Connect a broker when you need real positions, schedule repeated checks, and return to the conversation to update data, test assumptions, and review outcomes.
Use a desktop or mobile browser, or talk to your Agent through a connected [messaging channel](/concepts/channels).
## Decisions and responsibility
Analysis is informational and does not constitute investment advice. AI can misread data, miss information, or make calculation errors. Verify sources, timestamps, assumptions, and tool results before acting. Understand loss risks and authorization scope. Connector tools can run automatically in scheduled and messaging-channel workflows; see [permissions](/concepts/connectors#choose-a-permission-mode). Investment decisions and responsibility remain yours.
## Next steps
* [Compare Driven with general AI research tools](/why-driven/driven-vs-chatgpt-claude-gemini)
* [Quickstart](/get-started/quickstart)
* [Options](/concepts/options)
* [Plans and pricing](/get-started/plans-and-pricing)
# Build a Portfolio
Source: https://docs.driven.ai/guides/portfolio/build-a-portfolio
How to build a paper portfolio from scratch in Driven: translate a strategy into positions, size them with rules, and construct a diversified set of holdings.
This guide covers building a paper portfolio from scratch, going from a strategy to an actual set of positions. Whether you are testing a new approach or modeling a real allocation, Driven can help you translate a strategy into sized, diversified holdings.
The concept behind this is [Portfolio and Paper Trading](/concepts/portfolio-and-paper-trading), and it works best when your [Playbook](/concepts/playbook) holds your sizing rules.
## When to use this
* Starting a paper portfolio to test a strategy
* Modeling an allocation before committing real capital
* Building a diversified set of positions around a thesis
## Step 1: Define the strategy and constraints
Before buying anything, state what the portfolio is for and the rules it follows:
```text theme={null}
I want to build a paper portfolio of US large-cap quality companies. Maximum 10% per position, at least 8 names, diversified across sectors. Suggest a starting portfolio with the rationale for each holding.
```
If your [Playbook](/concepts/playbook) already holds your sizing and risk rules, the Agent will apply them automatically.
## Step 2: Review the proposed construction
Look at the whole before placing trades:
```text theme={null}
Before I build this, show me the proposed sector allocation and concentration, and flag any risks in the construction.
```
## Step 3: Build it
Once you are happy with the plan:
```text theme={null}
Build this portfolio in my paper account using the weights we discussed.
```
The Agent places the trades and the positions appear in the portfolio.
## Step 4: Document the thesis
Capture why you built it this way, so you can judge it later:
```text theme={null}
Summarize the thesis for this portfolio and the one metric to watch for each holding. Save it as a file.
```
## Step 5: Set up monitoring
A portfolio you do not watch drifts. Schedule a review:
```text theme={null}
Each week, review this portfolio's performance and flag any holding that has moved away from its thesis.
```
See [Portfolio health check](/guides/portfolio/portfolio-health-check) and [scheduled tasks](/concepts/scheduled-tasks).
## Common mistakes
* **Building without rules.** A portfolio with no sizing or diversification rules tends to become concentrated by accident. Set the constraints first.
* **No thesis on record.** If you do not write down why you bought, you cannot tell later whether the thesis broke or just the price moved.
## Related
* [Place a paper trade](/guides/portfolio/place-a-paper-trade) — the mechanics of trading
* [Portfolio health check](/guides/portfolio/portfolio-health-check) — review the portfolio
* [Build your Playbook](/guides/strategy/build-your-playbook) — where sizing rules live
* [Paper trading prompts](/prompts/paper-trading-prompts) — more templates
# Place a paper trade in Driven
Source: https://docs.driven.ai/guides/portfolio/place-a-paper-trade
Propose, review, and place a Driven paper trade. Select the correct account, check order status, and identify pending orders before cancellation.
This guide covers placing paper trades, simulated trades with real market data and virtual money. Paper trading lets you test ideas and practice without risking capital, and in Driven you place trades by describing them in plain language rather than filling out an order ticket.
The concept behind this is [Portfolio and Paper Trading](/concepts/portfolio-and-paper-trading).
Use the paper account for the supported market: USD accounts buy US stocks; HKD accounts buy Hong Kong stocks. Selling requires sufficient available holdings. Simulated matching does not reproduce real liquidity or slippage. See [paper trading limits](/concepts/portfolio-and-paper-trading#current-limitation).
## When to use this
* Testing a strategy before committing real money
* Tracking a research idea to see how it ages
* Practicing order workflows
* Comparing how different strategies perform
## Step 1: Describe the trade
Tell the Agent what you want in plain language:
```text theme={null}
Put 10% of available cash into NVDA in my paper account.
```
Or specify a dollar amount:
```text theme={null}
Propose a $5,000 paper purchase of [TICKER]. Show the account, quantity, concentration impact, and risks. Do not place an order.
```
## Step 2: Review the impact before it executes
A good habit is asking for the consequences before the trade goes through. The Agent can show how the trade changes your portfolio:
```text theme={null}
Show how this proposed trade affects concentration and risk. Do not submit an order until I give a separate instruction.
```
After reviewing the proposal, specify the paper account, symbol, quantity, and order type in a separate instruction. Check the resulting order status; a submitted order is not necessarily filled.
## Step 3: Use limit orders when you want a price
You are not limited to market orders. Set a price:
```text theme={null}
Place a limit order to buy [TICKER] at [price] in my paper account.
```
## Step 4: Manage open orders
List pending orders first. Confirm the portfolio and order IDs before asking the Agent to cancel them, especially when several orders share a ticker:
```text theme={null}
Cancel my open limit order on [TICKER].
```
```text theme={null}
Show me all my open paper orders and their status.
```
## Adjusting positions
Trimming and adding work the same way:
```text theme={null}
Cut my AAPL paper position in half and explain the impact on portfolio concentration.
```
```text theme={null}
Add to my [TICKER] position to bring it up to a 5% weight.
```
## Common mistakes
* **Trading without checking impact.** Always look at how a trade changes concentration and risk before executing.
* **Forgetting it is a test.** Paper trading is for learning. Use it to validate a process, then decide about real capital separately.
* **Ignoring coverage limits.** Paper trading coverage can be narrower than research coverage. Confirm the market is supported before building a workflow around it.
## Current limitation
Paper trading coverage may differ from research coverage. Check [supported markets](/reference/supported-markets) before relying on a trading workflow in a given market.
## Related
* [Portfolio and Paper Trading concept](/concepts/portfolio-and-paper-trading) — how it works
* [Build a portfolio](/guides/portfolio/build-a-portfolio) — construct a portfolio from scratch
* [Portfolio health check](/guides/portfolio/portfolio-health-check) — review what you've built
* [Paper trading prompts](/prompts/paper-trading-prompts) — more templates
# Portfolio Health Check
Source: https://docs.driven.ai/guides/portfolio/portfolio-health-check
How to run a portfolio health check in Driven: review concentration, sector exposure, risk, and holdings that have drifted from their thesis, and decide what to do.
This guide covers running a health check on your portfolio, a periodic review of concentration, exposure, and risk that catches problems before they cost you. The most common finding is hidden concentration: several positions quietly exposed to the same driver.
The Skill behind this workflow is [Portfolio Monitor](/skills/portfolio-monitor).
## When to use this
* Doing a periodic review of your holdings
* Checking whether you have become too concentrated
* Understanding your real risk and sector exposure
* Finding positions that no longer fit your strategy
## Step 1: Run the review
```text theme={null}
Review my portfolio. Check concentration, sector and risk exposure, diversification, and flag any holdings that need attention or no longer fit my strategy.
```
If your [Playbook](/concepts/playbook) holds your risk rules and sizing limits, the review will measure your holdings against them.
## Step 2: Surface hidden concentration
Position-level concentration is easy to see; driver-level concentration is not. Ask for it:
```text theme={null}
Map where my holdings overlap in their exposure. Are several positions exposed to the same theme, sector, or risk factor?
```
Owning five different names that all depend on the same driver is concentration, even if no single position is large.
## Step 3: Identify what needs action
```text theme={null}
Which holdings should I reconsider, and why? Separate "thesis has broken" from "just down in price."
```
A position being down is not a reason to act; a position whose thesis has broken is.
## Step 4: Decide on changes
```text theme={null}
If I wanted to reduce my biggest risk, what's the most efficient change to make, and what would it do to my concentration?
```
For acting on the review, see [Rebalancing strategies](/guides/portfolio/rebalancing-strategies).
## Make it recurring
A health check is most useful as a habit. Schedule it:
```text theme={null}
Each month, run a portfolio health check and flag concentration, risk, and any holdings that have drifted from their thesis.
```
## Common mistakes
* **Confusing price moves with broken theses.** A stock being down does not mean the investment case changed. Distinguish the two.
* **Only looking at position size.** The bigger risk is often correlated exposure across several positions, not one large holding.
* **Reviewing too rarely.** A portfolio drifts continuously. A periodic check catches drift before it becomes a problem.
## Related
* [Portfolio Monitor Skill](/skills/portfolio-monitor) — the workflow behind this guide
* [Risk exposure analysis](/guides/portfolio/risk-exposure-analysis) — deeper on risk
* [Rebalancing strategies](/guides/portfolio/rebalancing-strategies) — acting on the findings
# Rebalancing Strategies
Source: https://docs.driven.ai/guides/portfolio/rebalancing-strategies
How to rebalance a portfolio in Driven: bring positions back to target weights, trim winners, manage concentration, and model the impact before making changes.
This guide covers rebalancing, bringing a portfolio back in line with your target weights and risk after positions have drifted. Winners grow into oversized positions, losers shrink, and what was once balanced becomes concentrated. Rebalancing is how you reset, and Driven lets you model the impact before you act.
The Skill behind this workflow is [Portfolio Monitor](/skills/portfolio-monitor).
## When to use this
* Positions have drifted from their target weights
* A winner has grown into an oversized position
* Your portfolio has become more concentrated than intended
* You want to model changes before making them
## Step 1: See the drift
```text theme={null}
Show me how my current portfolio weights compare to my targets. Which positions have drifted most, and where am I now over- or under-weight?
```
## Step 2: Model the rebalance
Before trading, see what the changes would do:
```text theme={null}
Propose a rebalance back toward my targets. Show me what trades it would take and how the portfolio's concentration and risk would change afterward.
```
## Step 3: Weigh the trade-offs
Rebalancing is not free of judgment. Trimming a winner has costs:
```text theme={null}
If I trim my biggest winner back to target, what's the case for and against doing that right now? Is the position oversized on conviction or just on drift?
```
Sometimes an oversized winner reflects a strengthening thesis, not just price drift. Distinguish the two.
## Step 4: Execute in the paper account
```text theme={null}
Execute the rebalance in my paper account using the weights we agreed, and confirm the new allocation.
```
## Rebalancing approaches
There is no single right method. Common ones:
* **Calendar-based** — rebalance on a fixed schedule (quarterly, annually), regardless of drift
* **Threshold-based** — rebalance only when a position drifts beyond a set band (e.g. ±5% from target)
* **Opportunistic** — rebalance when adding new capital or when a thesis changes
Ask the Agent to help you reason about which fits your strategy.
## Common mistakes
* **Rebalancing too often.** Frequent rebalancing racks up friction and can cut winners short. Use bands or a schedule.
* **Trimming winners reflexively.** A position above target because the thesis is strengthening is different from one above target on a momentary pop.
* **Ignoring the tax and cost reality.** In a real account, rebalancing has costs. Model trades in paper, but remember the difference when applying to real capital.
## Related
* [Portfolio Monitor Skill](/skills/portfolio-monitor) — the workflow behind this guide
* [Portfolio health check](/guides/portfolio/portfolio-health-check) — find what needs rebalancing
* [Place a paper trade](/guides/portfolio/place-a-paper-trade) — execute the trades
# Risk Exposure Analysis
Source: https://docs.driven.ai/guides/portfolio/risk-exposure-analysis
How to analyze portfolio risk exposure in Driven: understand your real exposure to factors, sectors, and themes, and find correlated risks you didn't know you had.
This guide covers analyzing your portfolio's risk exposure, understanding what your holdings are really exposed to, beyond the obvious. The risks that hurt most are usually the ones you did not know you were carrying: correlated positions that all move together when one driver turns.
The Skill behind this workflow is [Portfolio Monitor](/skills/portfolio-monitor).
## When to use this
* Understanding what factors and themes your portfolio is exposed to
* Finding correlated risks across positions
* Stress-testing your holdings against a scenario
* Getting a clear picture of your real risk, not just position sizes
## Step 1: Map your exposures
```text theme={null}
Analyze my portfolio's risk exposure. Break down my exposure by sector, factor, and theme, and identify where my biggest concentrations of risk are.
```
## Step 2: Find correlated risk
The key insight is correlation, not just size:
```text theme={null}
Which of my holdings would move together in a downturn? Where am I exposed to the same underlying driver across multiple positions?
```
Three positions that all depend on rates staying low, or on one end market, are a single bet wearing three names.
## Step 3: Stress-test a scenario
```text theme={null}
If [scenario, e.g. rates rise sharply / this sector sells off], how would my portfolio likely behave, and which positions are most at risk?
```
## Step 4: Decide whether to adjust
```text theme={null}
Given this risk picture, what's the single change that would most reduce my exposure to [the biggest risk], and what would I give up?
```
## Common mistakes
* **Equating position size with risk.** A small position in a volatile, correlated name can carry more risk than a large position in a stable one.
* **Missing the common thread.** Diversification by name is not diversification by risk. Look for the shared driver.
* **Stress-testing only the obvious.** Test the scenario you are not worried about, not just the one you are.
## Prompt variations
```text theme={null}
What's my portfolio's exposure to [specific factor, e.g. interest rates, the dollar, a single end market]?
```
```text theme={null}
If my largest position fell 30%, what would happen to my overall portfolio, and is that an acceptable risk?
```
## Related
* [Portfolio Monitor Skill](/skills/portfolio-monitor) — the workflow behind this guide
* [Portfolio health check](/guides/portfolio/portfolio-health-check) — the broader review
* [Rebalancing strategies](/guides/portfolio/rebalancing-strategies) — acting on the analysis
# Analyze Before Earnings
Source: https://docs.driven.ai/guides/research/analyze-before-earnings
How to prepare for a company's earnings report in Driven: what to expect, the key numbers to watch, the setup, and how to read the result fast after it drops.
This guide covers preparing for an earnings report. Earnings are where theses get tested, and a little preparation turns a chaotic release into a clear read. Use this workflow in the days before a company reports, and again right after the numbers drop.
The Skill behind this workflow is [Stock Analysis](/skills/stock-analysis), focused on the earnings event.
## When to use this
* A company you follow reports soon
* You hold a position going into earnings
* You want to react quickly and correctly when numbers drop
## Step 1: Build the pre-earnings setup
A few days before the report, frame what matters:
```text theme={null}
[TICKER] reports earnings on [date]. Brief me before the report: what are the key numbers and metrics to watch, what is the consensus expecting, what guidance matters, and what's the setup into the print?
```
This tells you what to focus on so you are not reading the release cold.
## Step 2: Know what would move the stock
```text theme={null}
For [TICKER]'s upcoming earnings, what specific results would be bullish, what would be bearish, and what's already priced in?
```
Understanding what is already expected is what separates a real surprise from a non-event.
## Step 3: Read the result fast
When the numbers are out:
```text theme={null}
[TICKER] just reported. Summarize the results versus expectations, what guidance they gave, how the key metrics trended, and whether the thesis still holds.
```
Because the analysis cross-checks figures before presenting them, you get a read you can act on rather than a pile of raw numbers.
## Step 4: Update your view
```text theme={null}
Given these earnings, what changed in the investment case, and does my position still make sense?
```
## Common mistakes
* **Going in without a setup.** If you do not know what is expected, you cannot tell whether a result is good.
* **Reacting to headlines, not guidance.** The reported quarter often matters less than the forward guidance. Ask about both.
* **Forgetting your own thesis.** Tie the result back to why you owned (or wanted) the stock in the first place.
## Make it recurring
If you track an earnings-heavy watchlist, automate the calendar:
```text theme={null}
Each morning, tell me which of my watchlist names report earnings today or this week, and what to watch for each.
```
See [Earnings calendar alerts](/guides/scheduled/earnings-calendar-alerts) to set this up.
## Related
* [Stock Analysis Skill](/skills/stock-analysis) — the workflow behind this guide
* [Earnings calendar alerts](/guides/scheduled/earnings-calendar-alerts) — automate the calendar
* [Research a stock](/guides/research/research-a-stock) — the full single-stock workflow
# Compare Two Stocks
Source: https://docs.driven.ai/guides/research/compare-two-stocks
How to compare two stocks in Driven side by side: growth, margins, valuation, and competitive position, so you can decide which name is the stronger investment.
This guide covers comparing two (or more) stocks head to head. Use it when you have narrowed a decision to a few names and need to decide which one is the better investment, a question that is much easier to answer with the companies side by side than one at a time.
Use [Stock Analysis](/skills/stock-analysis) for a broad comparison of selected securities. Use [Competitor Analysis](/skills/competitor-analysis) when the objective is same-industry operating competitiveness.
## When to use this
* Choosing between two names in the same sector
* Deciding which competitor is winning
* Sanity-checking one company's numbers against a peer
## Step 1: Ask for a head-to-head comparison
```text theme={null}
Compare [TICKER A] and [TICKER B]. Cover relative growth, margins, valuation, competitive position, and which one looks like the stronger investment and why.
```
Example:
```text theme={null}
Compare NVDA and AMD. Who's winning on growth and margins, how do their valuations compare, and which has the better risk/reward right now?
```
## Step 2: Make the comparison concrete
A good comparison is specific about the gaps. If the answer is vague, push for the numbers:
```text theme={null}
Show the key metrics side by side in a table, then explain what the differences mean for the investment case.
```
## Step 3: Stress the weaker-looking name
It is easy to conclude the obvious winner is the winner. Test it:
```text theme={null}
Make the strongest case for [the name that looks weaker]. What would have to be true for it to be the better pick?
```
## Step 4: Decide and document
```text theme={null}
Based on this, which name fits a [your strategy] investor better, and what's the one metric to watch that would change the answer?
```
## Common mistakes
* **Comparing names that aren't really comparable.** Make sure the peer set makes sense; the Agent can help define it.
* **Letting valuation alone decide.** The cheaper stock is not automatically the better one. Weigh quality and trajectory too.
* **Ignoring the timeframe.** "Better investment" depends on horizon; state yours.
## Prompt variations
```text theme={null}
Compare [TICKER A], [TICKER B], and [TICKER C] and rank them for a long-term holder.
```
```text theme={null}
Compare [TICKER A] and [TICKER B] purely on capital efficiency and balance-sheet strength.
```
## Related
* [Competitor Analysis Skill](/skills/competitor-analysis) — the workflow behind this guide
* [Valuation Matrix](/skills/valuation-matrix) — value the winner properly
* [Research a stock](/guides/research/research-a-stock) — go deep on one name
# Macro Research
Source: https://docs.driven.ai/guides/research/macro-research
How to research macro themes in Driven: interest rates, geopolitics, and policy, and connect them to specific stocks and sectors that stand to benefit or suffer.
This guide covers researching macro themes, the big forces like interest rates, policy shifts, and geopolitics, and turning them into actionable investment ideas. Macro research is most useful when you connect the theme back to specific names, which is exactly what Driven is built to do.
The Skill behind this workflow is [Deep Research](/skills/deep-research).
## When to use this
* Understanding how a macro event affects markets
* Finding the companies that benefit or suffer from a policy change
* Building a view on a theme (rates, inflation, a geopolitical shift)
* Getting context before a major scheduled event (FOMC, key data releases)
## Step 1: Frame the theme
```text theme={null}
Run Deep Research on [macro theme]. Explain the dynamics, the likely paths, the key risks, and which markets and sectors are most exposed.
```
Example:
```text theme={null}
Run Deep Research on the impact of sustained higher interest rates on US equities. Which sectors are most exposed, and what's the transmission mechanism?
```
## Step 2: Connect macro to specific names
This is the step that turns a thesis into an idea:
```text theme={null}
Based on this theme, identify the public companies most positively and most negatively exposed, and explain the channel for each.
```
## Step 3: Stress-test the thesis
Macro views are easy to hold and hard to verify. Challenge it:
```text theme={null}
What's the strongest argument against this macro thesis, and what data would tell me I'm wrong?
```
## Step 4: Set up monitoring
Macro themes play out over time. Track them:
```text theme={null}
Set up a weekly summary of the key data and events relevant to [theme], and flag anything that changes the picture.
```
See [Macro event tracker](/guides/scheduled/macro-event-tracker) to automate this.
## Common mistakes
* **Stopping at the thesis.** A macro view with no named beneficiaries is an opinion, not an idea.
* **Ignoring what's priced in.** Markets often anticipate macro shifts; ask what is already reflected.
* **Treating macro as certain.** Frame it in probabilities and watch for disconfirming data.
## Prompt variations
```text theme={null}
How would [specific policy change] flow through to [sector]? Map the winners and losers.
```
```text theme={null}
What does the current rate environment mean for [specific stock or sector]?
```
## Related
* [Deep Research Skill](/skills/deep-research) — the workflow behind this guide
* [Macro event tracker](/guides/scheduled/macro-event-tracker) — monitor themes over time
* [Sector deep dive](/guides/research/sector-deep-dive) — drill into an exposed sector
# Research a Stock
Source: https://docs.driven.ai/guides/research/research-a-stock
A complete walkthrough for analyzing a single stock in Driven: the prompt, what to expect, how to pressure-test the answer, and follow-ups that sharpen the thesis.
This guide walks through researching one company end to end. Use it when you want a balanced, evidence-backed view of a single stock, the foundation for almost any investment decision.
The Skill behind this workflow is [Stock Analysis](/skills/stock-analysis). You do not need to name it; describing the task is enough.
## When to use this
* You are forming a thesis on a company
* You want to pressure-test a position you already hold
* You are preparing before earnings or before a decision
## Step 1: Ask for a structured analysis
Start with a prompt that asks for the full picture and, crucially, asks for evidence and risks:
```text theme={null}
Analyze [TICKER]. Cover business overview, recent developments, revenue and margin trends, valuation, technical context, sentiment, key risks, and what would change the conclusion.
```
If you care about specific drivers, name them. A focused prompt produces a sharper answer:
```text theme={null}
Analyze MSFT. Focus on Azure growth, AI capex, margin durability, valuation, and risks from cloud competition.
```
## Step 2: Run the review checklist
Before you trust the output, check it:
* Does the answer separate facts from assumptions?
* Are recent events dated?
* Are financial figures sourced?
* Are the risks concrete, not generic boilerplate?
* Does the conclusion explain what evidence would change it?
If any of these fail, do not start over — push back in the same conversation.
## Step 3: Pressure-test the thesis
The most valuable step is challenging the view you just got. Ask for the other side:
```text theme={null}
Compare this thesis against the bear case. What's the strongest argument that this is a bad investment?
```
A thesis that survives its own bear case is worth more than one that was never challenged.
## Step 4: Turn it into something you'll keep
When you are satisfied, capture it:
```text theme={null}
Turn this into a one-page investment memo.
```
For long research, ask the Agent to save it as a file in your workspace so you can revisit it later.
## Common mistakes
* **Asking too broadly.** "Tell me about NVDA" gets you a generic summary. Name what you care about.
* **Stopping at the bull case.** If you only read the optimistic view, you have confirmation, not research.
* **Not checking sources.** Always confirm the financial figures came from data, not recall. Ask the Agent to cite them.
## Prompt variations
```text theme={null}
Analyze [TICKER] strictly from a risk perspective. What could go wrong, and how likely is each?
```
```text theme={null}
Analyze [TICKER] as if I already own it at [price]. Should I add, hold, or trim, and why?
```
```text theme={null}
Give me the three things that matter most for [TICKER] over the next year, and how to monitor each.
```
## Related
* [Stock Analysis Skill](/skills/stock-analysis) — the workflow behind this guide
* [Analyze before earnings](/guides/research/analyze-before-earnings) — the earnings-specific version
* [Compare two stocks](/guides/research/compare-two-stocks) — when you're choosing between names
* [Stock research prompts](/prompts/stock-research-prompts) — more copyable prompts
# Run Deep Research
Source: https://docs.driven.ai/guides/research/run-deep-research
A full walkthrough of Driven's Deep Research workflow for multi-source investment research across companies, industries, and market themes, with exportable reports.
This guide covers running Deep Research, Driven's most thorough research workflow. Use it when a question needs evidence pulled from many sources and synthesized into a coherent, cited analysis, the kind of work that would take a human analyst hours.
The Skill behind this workflow is [Deep Research](/skills/deep-research).
## When to use this
* Company deep dives
* Industry maps and competitive landscapes
* Earnings-season summaries across multiple names
* Policy or macro event impact
* Theme research ("who benefits from X")
For a single metric or a quick market read, a lighter Skill is faster. Deep Research is for depth.
## Step 1: Define the scope clearly
The single biggest lever on Deep Research quality is a well-scoped question. Tell it what to include and exclude:
```text theme={null}
Run Deep Research on [topic]. Define the scope, collect evidence from multiple sources, separate facts from opinions, cite sources, summarize risks, and produce an analyst-style brief.
```
Example:
```text theme={null}
Run Deep Research on the AI data center power bottleneck. Identify public companies exposed to the theme, what evidence supports the thesis, and the biggest risks. Focus on US-listed names.
```
## Step 2: Let it work, then read critically
Deep Research collects evidence, checks key claims, and assembles a brief. Ask for source dates, assumptions, and unresolved gaps. Review the supporting evidence rather than treating a summary as complete verification.
When it finishes, read for evidence quality: are claims sourced, are figures cited, are the risks concrete?
## Step 3: Export the report
For anything long, save it:
```text theme={null}
Save the final report as a file in my workspace.
```
The report lands in the Agent's files, where you can revisit, share, or build on it.
## Step 4: Chain into specifics
Deep Research is often step one. Use its output to drive focused work:
```text theme={null}
From this research, take the three strongest names and run a full analysis and valuation on each.
```
## Common mistakes
* **Vague scope.** "Research AI" sprawls; "research US-listed AI data center infrastructure names and their power exposure" focuses.
* **Reading only the summary.** The value is in the evidence; skim the support, not just the conclusion.
* **Not exporting.** Long research you do not save is research you will redo.
## Prompt variations
```text theme={null}
Run Deep Research on [company] as a full investment case: business, financials, valuation, competition, and risks. Produce a downloadable memo.
```
```text theme={null}
Run Deep Research mapping the [industry] landscape: the layers, the key public companies at each, and where the value is accruing.
```
## Related
* [Deep Research Skill](/skills/deep-research) — the Skill reference
* [Sector deep dive](/guides/research/sector-deep-dive) — sector-focused research
* [Macro research](/guides/research/macro-research) — theme and policy research
# Sector Deep Dive
Source: https://docs.driven.ai/guides/research/sector-deep-dive
How to research an entire sector in Driven: map the landscape, find the leaders and laggards, understand the drivers, and build a shortlist of names worth analyzing.
This guide covers researching a whole sector rather than a single stock. Use it when you want to understand an industry, find the best names within it, or get oriented in a space before picking individual companies.
Use [Deep Research](/skills/deep-research) to map an industry and [Stock Screener](/skills/stock-screener) to discover companies. [Sector Radar](/skills/sector-radar) compares US sectors or industries, not companies inside one industry.
## When to use this
* Entering a new sector and wanting the lay of the land
* Finding the strongest companies within a theme
* Understanding what is driving a sector's moves
* Building a shortlist to research individually
## Step 1: Map the sector
Start with a scan to see the structure:
```text theme={null}
Research [industry] in [market]. Map its business segments, then use Stock Screener to find relevant companies. Show comparable metrics, sources, and gaps.
```
This gives you the shape of the sector: who is winning, who is lagging, and why.
## Step 2: Understand the drivers
Go a level deeper on what moves the whole group:
```text theme={null}
Do a deep dive on [sector]. What are the structural drivers, the key risks, the secular trends, and which companies are best positioned for them?
```
A sector's leaders only matter if you understand what they are leading on.
## Step 3: Build a shortlist
```text theme={null}
From this analysis, give me the 5 names most worth researching in detail, with the one reason each makes the shortlist.
```
## Step 4: Go deep on the best names
Hand the shortlist to single-stock workflows:
```text theme={null}
Analyze [top name from shortlist] in detail, focused on the sector drivers we identified.
```
## Common mistakes
* **Confusing the hot name with the best name.** The most-discussed company is not always the best investment in the sector.
* **Skipping the drivers.** If you do not understand what powers the sector, you cannot judge which companies are durably positioned.
* **Casting too wide.** Define the sector precisely; "AI" is not a sector, "AI data center infrastructure" is closer.
## Prompt variations
```text theme={null}
Map the [sector] supply chain and identify the public companies at each layer.
```
```text theme={null}
Which sub-segment of [sector] has the best fundamentals right now, and which names lead it?
```
## Related
* [Sector Radar Skill](/skills/sector-radar) — US sector and industry rotation
* [Deep Research Skill](/skills/deep-research) — the deeper research workflow
* [Macro research](/guides/research/macro-research) — zoom out one more level
* [Screen stocks by fundamentals](/guides/screening/screen-stocks-by-fundamentals) — filter within the sector
# Build multi-step scheduled tasks in Driven
Source: https://docs.driven.ai/guides/scheduled/advanced-scheduled-tasks
Build scheduled condition checks and multi-step research in Driven. Define data windows, delivery behavior, and external-tool execution limits.
Use scheduled tasks to check conditions and run a sequence of steps. The task wakes on its schedule, then evaluates available evidence; a condition in the prompt is not a real-time market-event subscription.
The concept behind this is [Scheduled Tasks](/concepts/scheduled-tasks).
Conditions in a prompt guide analysis; they are not notification switches. Scheduled conversation results may still be sent to a configured channel, including a no-events summary. Test runs and delivery; change task delivery settings if you do not want Telegram pushes.
## Conditional tasks
Every scheduled run checks the condition first. Ask for deeper analysis or notification only when it is met, then inspect run logs to confirm the behavior:
```text theme={null}
Every trading day, check my watchlist. If any name is down more than 8% from where I flagged it, run a quick analysis on what changed and whether the thesis still holds. If nothing crosses that line, report that no qualifying event was found.
```
The pattern is: check a condition, and act only if it is true. This is more useful than a task that runs the full analysis every day regardless.
## Multi-step tasks
A multi-step task does several things in sequence, producing a more complete output:
```text theme={null}
Every Monday morning: first, summarize how my portfolio performed last week. Then, check whether any holdings hit risk thresholds. Then, flag which holdings report earnings this week. Deliver as one combined briefing.
```
Each step builds on the last, and you get one coherent report instead of three separate alerts.
## Combining conditions and steps
The most powerful tasks combine both:
```text theme={null}
Each morning, scan my watchlist using the latest completed trading session. Flag daily price moves over 5% or volume above twice the preceding 20-session average. Check dated news and filings for possible explanations. Separate observed facts from hypotheses; do not infer institutional intent from price or volume alone. Summarize triggered names and remaining research questions. Do not trade.
```
## Tips for advanced tasks
* **State the condition precisely.** "If a name moves a lot" is vague; "if a name moves more than 5%" is actionable.
* **Order the steps logically.** Put the gating check first so the task does not do unnecessary work.
* **Keep output consolidated.** A single combined briefing is easier to act on than several fragmented alerts.
* **Test before relying.** Run the task a few times and refine the logic before trusting it for anything important.
## When to split into multiple tasks instead
Not everything belongs in one complex task. Split when:
* The pieces run on different schedules (hourly monitor vs weekly review)
* The pieces should go to different channels (urgent alerts to Telegram, deep reviews to web)
* One task is becoming hard to reason about
## Common mistakes
* **Over-engineering.** A task with too many branches is hard to debug. Start simple and add complexity only as needed.
* **Vague conditions.** Conditional logic needs concrete thresholds to work reliably.
* **Silent failures going unnoticed.** Review logs periodically so you know your tasks are running as intended.
## Related
* [Scheduled Tasks concept](/concepts/scheduled-tasks) — the fundamentals
* [Set up watchlist monitoring](/guides/scheduled/set-up-watchlist-monitoring) — conditional monitoring
* [Create a market briefing](/guides/scheduled/create-a-market-briefing) — the basic daily task
# Create a Market Briefing
Source: https://docs.driven.ai/guides/scheduled/create-a-market-briefing
How to set up a recurring pre-market briefing in Driven with a scheduled task: frequency, time zone, watchlist, format, and delivery to Telegram or the web.
This guide walks through building a recurring market briefing, a scheduled task that runs every morning and delivers a market read before you start your day. It is the single most popular scheduled task and a good first one to build.
This is a [scheduled task](/concepts/scheduled-tasks). The Agent uses data tools for routine briefings and chooses Skills only when the requested analysis needs them.
## When to use this
* You want a consistent market read every morning
* You track a watchlist and want overnight changes summarized
* You want the briefing waiting for you instead of building it yourself
## Step 1: Write the briefing prompt
A scheduled task is just a prompt that runs on a schedule. Start by writing the prompt as if you were asking it live, then add the scheduling details:
```text theme={null}
Every weekday at 8 AM New York time, scan my watchlist and send a pre-market briefing. Include major news, price moves, earnings events, and risks to watch today.
```
## Step 2: Specify the six elements
A complete scheduled task includes:
* **Frequency** — "every weekday"
* **Time zone** — "8 AM New York time" (always anchor the zone)
* **Universe** — "my watchlist" (or name specific tickers/markets)
* **Output format** — "a briefing with major news, price moves, earnings, risks"
* **Delivery channel** — where it should arrive (web, Telegram)
* **Thresholds** — what to emphasize or flag
The more of these you specify, the more useful and less noisy the output.
## Step 3: Choose where it arrives
For a morning briefing you read on your phone, route it to Telegram:
```text theme={null}
Deliver the briefing to Telegram each morning.
```
For a briefing you want to inspect closely with files, keep it in the web app.
## Step 4: Tune it after a few runs
Let it run for a few days, then adjust. If it is too broad, narrow the universe or tighten the format:
```text theme={null}
Make the briefing shorter: top 5 things only, ranked by importance, one line each.
```
If it is missing something, add it:
```text theme={null}
Also include notable options activity from available US and Hong Kong equity and ETF options data on my watchlist names.
```
## Example variations
A focused, scannable version:
```text theme={null}
Every weekday at 7:30 AM New York time, give me a 5-bullet pre-market briefing on my watchlist: biggest move, key news, any earnings today, one risk to watch, and one opportunity. Deliver to Telegram.
```
A market-wide version:
```text theme={null}
Every trading day at 8 AM, summarize the overnight market action: index moves, notable sector moves, the main news driving them, and what to watch at the open.
```
## Common mistakes
* **No time zone.** "8 AM" is ambiguous. Always anchor it.
* **Too broad.** A briefing covering everything covers nothing well. Scope it to your watchlist or market.
* **Set and forget.** Review the first few runs and tune. The best briefings are iterated, not written once.
## Related
* [Scheduled Tasks concept](/concepts/scheduled-tasks) — how scheduled tasks work
* [Market Pulse Skill](/skills/market-pulse) — macro and event-transmission analysis
* [Set up watchlist monitoring](/guides/scheduled/set-up-watchlist-monitoring) — threshold-based alerts
* [Scheduled task prompts](/prompts/scheduled-task-prompts) — more templates
# Earnings Calendar Alerts
Source: https://docs.driven.ai/guides/scheduled/earnings-calendar-alerts
How to set up earnings calendar alerts in Driven, so you know which of your watchlist names report each day or week and what to watch for in each report.
Use a scheduled check to track upcoming earnings for your holdings or watchlist. Dates and release times can change, so include source dates and missing information in the report.
The concept behind this is [Scheduled Tasks](/concepts/scheduled-tasks), and it pairs with the [Analyze before earnings](/guides/research/analyze-before-earnings) workflow.
## When to use this
* You hold or follow names that report quarterly and do not want to miss a date
* You want a heads-up on what matters before each report
* Earnings season tends to sneak up on you
## Step 1: Set up the daily calendar check
```text theme={null}
Each morning, tell me which of my watchlist names report earnings today or this week. For each, note the date, the time (before or after market), and the key things to watch.
```
## Step 2: Add the pre-earnings context
Make the alert useful, not just a date:
```text theme={null}
For any name reporting in the next two days, include what consensus expects, what guidance matters, and what result would be bullish or bearish.
```
This gives you the setup ahead of time so you are not reading the release cold. See [Analyze before earnings](/guides/research/analyze-before-earnings) for the full pre-earnings workflow.
## Step 3: Get the post-report read
You can also schedule a follow-up that summarizes results on report days:
```text theme={null}
On any day a watchlist name reports, send me a summary after the release: results versus expectations, guidance, key metric trends, and whether the thesis still holds.
```
## Step 4: Choose delivery
Route the calendar to wherever you plan your day. Telegram works well for a morning heads-up; the web app is better if you want to dig into the pre-earnings analysis right away.
## Example variations
A simple calendar:
```text theme={null}
Every morning, list my watchlist names reporting earnings this week with dates and times. Deliver to Telegram.
```
A calendar with prep built in:
```text theme={null}
Each weekday morning, tell me which of [tickers] report in the next three days, what to watch for each, and what's already priced in.
```
## Common mistakes
* **Date only, no context.** Knowing the date is half the value. Ask what to watch.
* **No post-report follow-up.** The alert that matters most is often the read right after the numbers drop.
* **Watchlist out of date.** Keep your watchlist current so the calendar covers what you actually follow.
## Related
* [Analyze before earnings](/guides/research/analyze-before-earnings) — the full pre-earnings workflow
* [Scheduled Tasks concept](/concepts/scheduled-tasks) — how scheduled tasks work
* [Scheduled task prompts](/prompts/scheduled-task-prompts) — more templates
# Macro Event Tracker
Source: https://docs.driven.ai/guides/scheduled/macro-event-tracker
How to set up a macro event tracker in Driven to stay ahead of FOMC meetings, economic data, and policy events, and understand their impact on your holdings.
This guide covers tracking macro events with a scheduled task, so you stay ahead of the calendar of central bank meetings, economic releases, and policy events that move markets. The value is not just knowing an event is coming; it is knowing what it means for the names you hold.
The concept behind this is [Scheduled Tasks](/concepts/scheduled-tasks), and it pairs with [Macro research](/guides/research/macro-research).
## When to use this
* You want a heads-up before major macro events (FOMC, jobs data, CPI)
* You hold positions sensitive to rates, inflation, or policy
* You want macro context delivered on a schedule rather than chasing it
## Step 1: Track the calendar
```text theme={null}
Each week, give me the key macro events coming up: central bank meetings, major economic releases, and policy events. For each, note the date, what's expected, and why it matters for markets.
```
## Step 2: Connect events to your exposure
Generic macro calendars are everywhere. The useful version ties events to your holdings:
```text theme={null}
For each upcoming macro event, flag which of my holdings or watchlist names are most exposed and in which direction.
```
## Step 3: Get the pre-event setup
```text theme={null}
Before [recurring event, e.g. each FOMC meeting], brief me on what's expected, what's priced in, and the scenarios for the market reaction.
```
## Step 4: Get the post-event read
```text theme={null}
After the event, summarize what happened versus expectations, how markets reacted, and what it means going forward for my exposure.
```
## Example variations
A weekly macro calendar:
```text theme={null}
Every Sunday evening, give me the week's key macro events with dates, expectations, and the main names or sectors exposed to each. Deliver to Telegram.
```
An event-specific tracker:
```text theme={null}
Track the path of interest rates. Each week, summarize any new data or commentary that changes the outlook, and flag the impact on my rate-sensitive holdings.
```
## Common mistakes
* **Calendar without context.** A list of dates is not useful unless it connects to your portfolio.
* **Ignoring what's priced in.** Markets anticipate macro events; the reaction depends on the surprise, not the event itself. Ask what is already reflected.
* **Treating macro as certain.** Frame events in scenarios and probabilities, not single predictions.
## Related
* [Macro research](/guides/research/macro-research) — the full macro research workflow
* [Scheduled Tasks concept](/concepts/scheduled-tasks) — how scheduled tasks work
* [Scheduled task prompts](/prompts/scheduled-task-prompts) — more templates
# Schedule watchlist monitoring in Driven
Source: https://docs.driven.ai/guides/scheduled/set-up-watchlist-monitoring
Schedule watchlist checks in Driven with explicit thresholds and data windows. Verify time zones, Telegram delivery, and no-events notifications.
Use a scheduled task to check your watchlist at specified times. Define thresholds and comparison windows so the report focuses on meaningful changes; the task does not subscribe to every market event.
The concept behind this is [Scheduled Tasks](/concepts/scheduled-tasks).
Conditions in a prompt guide analysis; they are not notification switches. Scheduled conversation results may still be sent to a configured channel, including a no-events summary. Test runs and delivery; change task delivery settings if you do not want Telegram pushes.
## When to use this
* You track a set of names and want to be alerted to changes
* You want signal, not a daily report of nothing happening
* You want to catch big moves, news, or level breaks without watching the screen
## Step 1: Define what counts as notable
A monitor is only as good as its trigger. Decide what is worth an alert:
```text theme={null}
Monitor my watchlist throughout the trading day. Alert me only when a name moves more than 5%, has significant news, or breaks a key technical level.
```
Notice the word "only." That is what keeps the monitor from becoming noise.
## Step 2: Set the cadence and channel
```text theme={null}
Check every hour during market hours and send any alerts to Telegram. If nothing crosses the thresholds, report that no qualifying event was found.
```
A monitor that says "nothing happened" every hour trains you to ignore it. Let silence mean "all clear."
## Step 3: Make alerts actionable
An alert should tell you enough to decide whether to act:
```text theme={null}
When you alert me, include what happened, why it likely moved, and whether it changes anything for that name.
```
## Step 4: Layer monitors if needed
You can run more than one monitor at different sensitivities:
* A fast, high-threshold monitor for big intraday moves (Telegram)
* A slower, end-of-day monitor summarizing everything notable (web app)
## Example variations
A move-and-news monitor:
```text theme={null}
During market hours, watch [tickers]. Alert me to any move over 4%, any material news, or any unusual volume. If no event qualifies, report that briefly. Deliver to Telegram.
```
A level-break monitor:
```text theme={null}
Watch [ticker]. Alert me if it breaks above [level] or below [level], with the context for the move.
```
## Common mistakes
* **No threshold.** A monitor with no trigger is just a recurring report. Define what is worth an alert.
* **Threshold too low.** A 1% trigger fires constantly. Set it where a move actually matters to you.
* **Alerts with no context.** "NVDA -6%" is incomplete. Ask for the why and the so-what.
## Related
* [Scheduled Tasks concept](/concepts/scheduled-tasks) — how scheduled tasks work
* [Unusual volume monitor](/guides/scheduled/unusual-volume-monitor) — a volume-specific monitor
* [Create a market briefing](/guides/scheduled/create-a-market-briefing) — the daily-summary counterpart
* [Scheduled task prompts](/prompts/scheduled-task-prompts) — more templates
# Unusual Volume Monitor
Source: https://docs.driven.ai/guides/scheduled/unusual-volume-monitor
How to set up an unusual volume monitor in Driven to catch names trading on abnormal volume, often an early sign that something is happening before the news is obvious.
This guide covers monitoring for unusual volume — abnormal trading activity that often precedes or accompanies a meaningful move. Volume spikes are one of the earliest signs that something is happening in a name, sometimes before the reason is widely known.
A volume monitor is a specialized watchlist monitor focused on a single dimension. For the general mechanics — thresholds, cadence, delivery channel, and making alerts actionable — follow [Set up watchlist monitoring](/guides/scheduled/set-up-watchlist-monitoring). This guide covers only what is specific to volume. The underlying concept is [Scheduled Tasks](/concepts/scheduled-tasks).
## When to use this
* You want early warning when a name starts trading abnormally
* You track names where a volume spike would prompt you to dig in
* You want to catch developing situations before they hit the headlines
## Define "unusual" against a baseline
Volume only means something relative to a name's normal. Frame the trigger as a multiple of average, not a raw number:
```text theme={null}
Monitor my watchlist for unusual volume. Alert me when a name trades significantly above its average volume — for example, more than 2x — and tell me what's likely driving it.
```
This is the one setting that makes a volume monitor work. A fixed share count fires constantly on liquid names and never on thin ones; a multiple of average normalizes across the whole list.
## Read volume together with price and flow
Raw volume is only half the signal. A spike is a question — pair it with the cause and the direction:
```text theme={null}
When you flag unusual volume, check dated news, filings, and events for possible explanations. Separate observations from hypotheses. Do not infer institutional intent or trade direction from volume alone; identify missing evidence.
```
Check whether the market provides relevant flow or disclosure evidence. Volume alone does not identify who traded or why. US 13F holdings are delayed disclosures; A-share Dragon & Tiger lists cover a specific disclosed population. Use [Insider Tracker](/skills/smart-money) for supported disclosures and retain their dates and scope.
## Example variations
A watchlist volume monitor:
```text theme={null}
During market hours, watch [tickers] for volume more than 2x their average. Flag each with the likely cause and whether it's buying or selling pressure. Deliver to Telegram.
```
A broader scan:
```text theme={null}
Each day after market close, tell me which names in [sector or universe] had the most unusual volume today and what drove it.
```
For cadence, delivery, and threshold checks, reuse the patterns in [Set up watchlist monitoring](/guides/scheduled/set-up-watchlist-monitoring).
## Common mistakes
* **No baseline.** "High volume" means nothing without an average to compare against. Frame it as a multiple of normal.
* **Spikes without explanation.** Always have the monitor look for the cause; raw volume is only half the signal.
* **Acting on the spike alone.** Volume tells you to look; it does not tell you to buy or sell.
## Related
* [Set up watchlist monitoring](/guides/scheduled/set-up-watchlist-monitoring) — the general monitoring mechanics this guide builds on
* [Insider Tracker](/skills/smart-money) — inspect dated transaction and ownership disclosures
* [Scheduled task prompts](/prompts/scheduled-task-prompts) — more templates
# ETF Flow Analysis
Source: https://docs.driven.ai/guides/screening/etf-flow-analysis
How to analyze ETF and fund flows in Driven: see where money is moving across sectors and themes, and use flows to read positioning and sentiment.
This guide covers analyzing ETF and fund flows, where money is moving at the sector and theme level. Flows reveal positioning that individual stock data does not: which themes are attracting capital and which are being abandoned.
Use [Stock Analysis](/skills/stock-analysis) for named funds and inspect the available ETF data. Direct fund-flow coverage varies by source. Prices, assets under management, and holdings changes are not interchangeable with net subscriptions and redemptions.
## When to use this
* Seeing which sectors or themes are attracting or losing capital
* Reading positioning and sentiment at a higher level than single stocks
* Confirming a sector thesis with flow data
* Spotting rotation between themes
## Step 1: Look at the flow picture
```text theme={null}
Show available ETF net-flow evidence for [sector or theme]. State the market, dates, units, and source. If direct flow data is unavailable, explain the gap and do not substitute price performance.
```
Example:
```text theme={null}
What do ETF flows tell us about positioning in semiconductors and AI infrastructure right now?
```
## Step 2: Read rotation
Flows are most telling when you compare across themes:
```text theme={null}
Which sectors are seeing the strongest inflows and outflows recently, and is there a rotation pattern?
```
## Step 3: Connect flows to names
```text theme={null}
For the theme attracting the most flows, which individual names are the main beneficiaries, and are fundamentals keeping up with the money?
```
## Step 4: Use flows as context, not a trigger
```text theme={null}
Given these flows and the fundamentals, is the capital movement justified or is it a crowding risk?
```
## Common mistakes
* **Chasing flows.** Heavy inflows can signal crowding, which is a risk, not just a green light.
* **Ignoring fundamentals.** Money flowing into a theme does not make the underlying businesses good.
* **Reading flows in isolation.** Combine with positioning and fundamentals for a complete picture.
## Prompt variations
```text theme={null}
Are flows into [theme] accelerating or decelerating, and what's the trend over the past few months?
```
```text theme={null}
Compare flows into [sector A] versus [sector B] and what the divergence suggests.
```
## Related
* [Stock Analysis](/skills/stock-analysis) — named ETF and fund analysis
* [Sector Radar](/skills/sector-radar) — US sector and industry rotation
* [Data and coverage](/concepts/data-and-coverage) — the ETF and fund data behind this guide
# Find Momentum Plays
Source: https://docs.driven.ai/guides/screening/find-momentum-plays
How to screen for momentum and technical setups in Driven: relative strength, breakouts, and volume, combined with fundamentals to avoid chasing weak names.
This guide covers screening for momentum, finding stocks with strong price action and technical setups. Momentum screening surfaces names that are moving; combining it with a fundamental check keeps you from chasing names that are moving for no good reason.
The Skill behind this workflow is [Stock Screener](/skills/stock-screener).
## When to use this
* Looking for stocks with strong relative strength
* Finding technical breakouts or setups
* Spotting names with unusual volume or accelerating moves
* Adding a momentum overlay to a fundamental shortlist
## Step 1: Screen for the technical setup
```text theme={null}
Find [market] stocks with strong relative strength, trading near [setup, e.g. 52-week highs], with above-average volume. Show the technical context for each.
```
Example:
```text theme={null}
Find US stocks breaking out to new highs with strong relative strength and rising volume over the past month.
```
## Step 2: Add a fundamental filter
This is what separates a momentum strategy from chasing. Layer in quality:
```text theme={null}
From that list, keep only the names that also have improving fundamentals, positive revenue growth, and reasonable balance sheets.
```
A stock with strong momentum and improving fundamentals is a very different proposition from one with momentum alone.
## Step 3: Check the positioning
```text theme={null}
For the top names, what's the capital flow picture? Are institutions accumulating, or is this a retail-driven move?
```
For A-shares, this draws on Driven's flow data, including Dragon & Tiger lists and institutional-versus-retail flow.
## Step 4: Define your risk
Momentum cuts both ways. Before acting:
```text theme={null}
For [name], where's the technical level that would invalidate the setup, and what's the risk if the momentum reverses?
```
## Common mistakes
* **Chasing momentum with no fundamentals.** Price strength alone is fragile. Always add a quality check.
* **Ignoring volume.** A move on thin volume is less reliable than one confirmed by participation.
* **No exit plan.** Momentum trades need a defined invalidation level. Set it before you enter.
## Prompt variations
```text theme={null}
Find [market] stocks that just had a volume spike with a positive catalyst.
```
```text theme={null}
Screen for names with improving momentum off a base, not extended ones that have already run.
```
## Related
* [Stock Screener Skill](/skills/stock-screener) — the workflow behind this guide
* [Screen stocks by fundamentals](/guides/screening/screen-stocks-by-fundamentals) — the fundamental counterpart
* [Insider Tracker](/skills/smart-money) — check who's behind the move
# Insider Trading Signals
Source: https://docs.driven.ai/guides/screening/insider-trading-signals
How to read insider buying and selling in Driven: distinguish meaningful signals from noise, weigh cluster buys, and use insider activity to inform a thesis.
This guide covers reading insider activity, the buying and selling by a company's own executives and directors. Insider behavior can be a useful signal because insiders know their business better than anyone, but it has to be read carefully: not every transaction means what it seems to.
The Skill behind this workflow is [Insider Tracker](/skills/smart-money).
This guide is about reading legally disclosed insider transactions as an investment signal. It is not about trading on material non-public information, which is illegal. Driven works only with publicly disclosed filings.
## When to use this
* Checking whether insiders are buying or selling a name
* Reading insider activity as a conviction signal
* Adding an insider check to a thesis
* Spotting unusual insider behavior
## Step 1: Pull the insider activity
```text theme={null}
Run Insider Tracker on [TICKER] focused on insider activity. Show recent insider buys and sells, who transacted, the size, and the context.
```
Example:
```text theme={null}
Show me recent insider activity for [TICKER]. Are executives buying or selling, and is there anything unusual?
```
## Step 2: Separate signal from noise
Not all insider trades carry the same weight. Ask the Agent to help interpret:
```text theme={null}
For this insider activity, what's likely meaningful versus routine? Distinguish open-market purchases from scheduled sales and option exercises.
```
Open-market purchases by senior executives tend to carry more signal than routine sales, which often reflect diversification or tax planning rather than a view on the stock.
## Step 3: Look for clusters
One insider buying is interesting; several buying together is more so:
```text theme={null}
Is there any cluster of insider buying at [TICKER], where multiple insiders bought around the same time?
```
## Step 4: Tie it to the thesis
```text theme={null}
Does the insider behavior support or contradict the investment case for [TICKER], and how much weight should it carry?
```
## Common mistakes
* **Reading every sale as bearish.** Insiders sell for many reasons unrelated to the outlook. Sales are weak signals.
* **Overweighting one transaction.** A single trade is thin evidence. Clusters and patterns matter more.
* **Forgetting the lag.** Insider transactions are disclosed after the fact; note the dates.
## Prompt variations
```text theme={null}
Find names in my watchlist with notable insider buying recently.
```
```text theme={null}
Show insider sentiment for [sector]: where are insiders net buyers versus net sellers?
```
## Related
* [Insider Tracker Skill](/skills/smart-money) — the workflow behind this guide
* [Track institutional holdings](/guides/screening/track-institutional-holdings) — the institutional counterpart
* [Data and coverage](/concepts/data-and-coverage) — the insider data behind this guide
# Screen Stocks by Fundamentals
Source: https://docs.driven.ai/guides/screening/screen-stocks-by-fundamentals
How to screen for stocks by fundamentals in Driven: describe your criteria in plain language and get a ranked shortlist of companies that match, with the metrics that matter.
This guide covers fundamental screening, finding stocks that match financial criteria you care about. Instead of clicking through filter menus, you describe what you are looking for in plain language and Driven returns the names that fit.
The Skill behind this workflow is [Stock Screener](/skills/stock-screener).
## When to use this
* Finding investment candidates that meet specific financial criteria
* Narrowing a large universe to a workable shortlist
* Discovering quality companies you would not have thought to look at
## Step 1: Describe your criteria
Tell the Agent what a candidate looks like to you:
```text theme={null}
Find profitable [market] companies with revenue growth above [X]%, positive free cash flow, [margin criteria], and reasonable valuation. Rank the results and show the key metrics for each.
```
Example:
```text theme={null}
Find profitable US software companies with revenue growth above 15%, positive free cash flow, gross margins above 70%, and reasonable valuation.
```
## Step 2: Read the shape of the results
Look at what comes back before tightening. If you get 50 names, the screen is too loose; if you get zero, it is too tight. Adjust the binding constraint:
```text theme={null}
That returned too many names. Add a constraint for return on invested capital above 15% and re-rank.
```
## Step 3: Hand the shortlist to deeper analysis
A screen produces candidates, not conclusions. Take the best names deeper:
```text theme={null}
Take the top 3 from this screen and run a full analysis on each, focused on whether the growth is durable.
```
## Common mistakes
* **Over-filtering.** Stacking too many constraints can leave you with nothing or with statistical flukes. Start loose.
* **Screening on one factor.** A single-metric screen ("cheap stocks") surfaces value traps. Combine quality with valuation.
* **Trusting the screen as the answer.** A screen is a starting list. The research happens after.
## Prompt variations
```text theme={null}
Screen for [market] dividend growers: rising dividends, payout ratio under 60%, and stable free cash flow.
```
```text theme={null}
Find [market] companies with improving margins and accelerating revenue over the last four quarters.
```
```text theme={null}
Screen A-share companies with strong fundamentals and net institutional inflows over the past month.
```
## Related
* [Stock Screener Skill](/skills/stock-screener) — the workflow behind this guide
* [Find momentum plays](/guides/screening/find-momentum-plays) — technical-factor screening
* [Screening prompts](/prompts/screening-prompts) — more copyable prompts
# Track Institutional Holdings
Source: https://docs.driven.ai/guides/screening/track-institutional-holdings
How to track institutional holdings in Driven: see what funds are accumulating or exiting, read changes over time, and use positioning to confirm or challenge a thesis.
This guide covers tracking what institutions own and how their positions are changing. Institutional positioning is one of the more useful signals available to an individual investor, the accumulation and distribution that often precedes a story becoming widely known.
The Skill behind this workflow is [Insider Tracker](/skills/smart-money).
## When to use this
* Checking whether institutions are accumulating or exiting a name
* Understanding who owns a stock and how conviction is changing
* Confirming or challenging a thesis with positioning data
* Spotting names where smart money is moving early
## Step 1: Pull the holdings picture
```text theme={null}
Run Insider Tracker on [TICKER]. Show institutional holdings, recent changes, who the largest holders are, and whether positions are growing or shrinking.
```
Example:
```text theme={null}
What are institutions doing with NVDA? Show holdings changes over the last few quarters and whether conviction is rising or falling.
```
## Step 2: Read the direction, not just the level
A large institutional stake is less interesting than a changing one. Focus on the trend:
```text theme={null}
Is the institutional ownership of [TICKER] trending up or down, and which holders are driving the change?
```
## Step 3: Mind the reporting lag
Holdings data is reported periodically, so it reflects positions as of a past date. Ask the Agent to be explicit:
```text theme={null}
What are the as-of dates for this holdings data, and how stale might the picture be?
```
## Step 4: Combine with your own view
Positioning is an input, not a verdict:
```text theme={null}
Given this positioning and the fundamentals, does the institutional behavior support or contradict the investment case?
```
## Common mistakes
* **Treating holdings as real-time.** The data lags. Always note the as-of date.
* **Following blindly.** Institutions are wrong too. Use positioning to inform, not to decide.
* **Ignoring why.** A fund exiting for portfolio reasons is different from one exiting on a thesis change; the level alone does not tell you which.
## Prompt variations
```text theme={null}
Which institutions added to [sector] most aggressively last quarter, and what does that suggest?
```
```text theme={null}
Show me names in my watchlist with the biggest positive change in institutional ownership recently.
```
## Related
* [Insider Tracker Skill](/skills/smart-money) — the workflow behind this guide
* [Insider trading signals](/guides/screening/insider-trading-signals) — the insider counterpart
* [Data and coverage](/concepts/data-and-coverage) — the holdings data behind this guide
# Build your Playbook
Source: https://docs.driven.ai/guides/strategy/build-your-playbook
Define the rules for a specific portfolio and keep them separate from Agent-wide preferences.
A [Playbook](/concepts/playbook) stores the strategy for **one portfolio**. Use it for the investment universe, watchlist, risk limits, sizing, benchmark, and entry or exit discipline.
Use the Agent's Profile for language, tone, and general research preferences. Memory stores facts, research notes, and references. A Playbook guides analysis; it is not a hard risk-control system and does not trigger real trades by itself.
## 1. Choose the portfolio
Create or select the paper portfolio you want to manage. If the Agent has several, name the portfolio in every update so its rules are unambiguous.
## 2. Define the strategy
```text theme={null}
Update the Playbook for my [portfolio name] paper portfolio.
Universe: profitable US and Hong Kong software companies.
Focus: durable revenue, cash generation, valuation, and balance-sheet strength.
Risk: maximum 10% in one position; flag concentrations above 15%.
Watchlist: AAPL, MSFT, NVDA, and TSM.
Show proposed rule changes before saving them.
```
Add a benchmark, review cadence, and the evidence that would invalidate a holding. State how to handle conflicts between rules.
## 3. Keep Agent-wide preferences separate
```text theme={null}
Save to this Agent's Profile: answer concisely, separate facts from assumptions,
include source dates, and explain the main risks.
```
## 4. Read back and revise
Ask the Agent to show the chosen portfolio's Playbook. Check that it saved the correct portfolio, contains consistent rules, and reflects your current risk preferences. Revisit it when your strategy or assumptions change.
One-off research belongs in chat. Recurring checks belong in [scheduled tasks](/concepts/scheduled-tasks). Before any trade, verify how the proposed action fits the rules and the account's capabilities.
## Related
* [Playbook](/concepts/playbook)
* [Playbook templates](/prompts/playbook-templates)
* [Quickstart](/get-started/quickstart)
# Growth Investing Setup
Source: https://docs.driven.ai/guides/strategy/growth-investing-setup
How to configure Driven for growth investing: a Playbook, screens, and research workflows tuned for finding durable, high-growth companies with strong unit economics.
This guide is a worked example: configuring a Driven Agent for a growth investing strategy. Growth investing has different priorities from value — durability of growth and the size of the opportunity matter more than buying below intrinsic value — and the setup reflects that.
For other styles, see [Value investing setup](/guides/strategy/value-investing-setup) and [Macro trading setup](/guides/strategy/macro-trading-setup).
## Step 1: Create a dedicated Agent
```text theme={null}
Create an agent named "Growth Research" for a growth investing strategy.
```
## Step 2: Build the Playbook
```text theme={null}
Add to this portfolio's Playbook:
Philosophy: Growth investing. Own durable, high-growth companies expanding into large markets, and hold through volatility as long as the growth thesis holds.
Universe: US and Hong Kong listed growth companies, primarily technology, software, and emerging category leaders.
What I look for: high and durable revenue growth, large and expanding addressable markets, improving unit economics, strong competitive positioning, and a path to (or existing) profitability and free cash flow.
What I avoid: growth that is decelerating without a reacceleration path, businesses burning cash with no route to profitability, and companies whose growth depends on a single fragile driver.
Risk and sizing: maximum 10% per position, willing to tolerate volatility, but exit if the growth thesis breaks.
Research style: prioritize revenue growth durability, market size, unit economics, competitive moats, and the trajectory toward profitability. Be explicit about valuation risk given growth premiums, and state what would break the growth thesis.
```
## Step 3: Set up growth screening
```text theme={null}
Screen for US and Hong Kong companies with revenue growth above 20%, improving margins or a clear path to profitability, strong competitive positioning, and large addressable markets. Rank by the durability of the growth.
```
See [Screen stocks by fundamentals](/guides/screening/screen-stocks-by-fundamentals).
## Step 4: Research candidates the growth way
```text theme={null}
Analyze [TICKER] as a growth investment. Cover revenue growth and its durability, the size and expansion of the addressable market, unit economics, competitive positioning, the path to profitability and free cash flow, valuation risk, and what would break the growth thesis.
```
For high-growth names, valuation needs care because of the premium:
```text theme={null}
Run a Valuation Matrix on [TICKER]. Account for the growth profile, be explicit about the assumptions baked into the valuation, and tell me what growth and margin path the current price implies.
```
## Step 5: Monitor the growth thesis
For growth, the monitoring centers on whether growth is holding:
```text theme={null}
Each quarter, review my growth holdings. For each, check whether revenue growth is holding, decelerating, or reaccelerating, whether unit economics are improving, and flag any where the growth thesis is weakening.
```
The single most important monitoring question for a growth holding is: is the growth still there?
## Step 6: Watch the deceleration risk
```text theme={null}
Alert me if any of my growth holdings shows signs of decelerating growth, margin deterioration, or a weakening competitive position.
```
## Common mistakes
* **Ignoring valuation entirely.** Growth justifies a premium, not any price. Always check what the valuation implies.
* **Holding through a broken thesis.** Growth investing tolerates volatility, but not a genuine breakdown in growth. Distinguish a dip from a deceleration.
* **Chasing the story.** A compelling narrative is not the same as durable growth with sound unit economics. The Playbook keeps the focus on fundamentals.
## Related
* [Build your Playbook](/guides/strategy/build-your-playbook) — the Playbook fundamentals
* [Valuation Matrix](/skills/valuation-matrix) — valuing growth companies
* [Screen stocks by fundamentals](/guides/screening/screen-stocks-by-fundamentals) — finding candidates
* [Value investing setup](/guides/strategy/value-investing-setup) — the value counterpart
# Macro Trading Setup
Source: https://docs.driven.ai/guides/strategy/macro-trading-setup
How to configure Driven for macro and thematic trading: a Playbook, event tracking, and research workflows tuned to top-down themes, rates, and policy-driven opportunities.
This guide is a worked example: configuring a Driven Agent for a macro and thematic trading strategy. Macro trading is top-down: it starts from a view on rates, policy, or a structural theme and works down to the names that express it, which is the opposite direction from bottom-up stock picking.
For other styles, see [Value investing setup](/guides/strategy/value-investing-setup) and [Growth investing setup](/guides/strategy/growth-investing-setup).
## Step 1: Create a dedicated Agent
```text theme={null}
Create an agent named "Macro Trading" for a top-down macro and thematic strategy.
```
## Step 2: Build the Playbook
```text theme={null}
Add to this portfolio's Playbook:
Philosophy: Top-down macro and thematic trading. Start from a macro or structural view, then find the cleanest ways to express it. Hold while the thesis and the data support it.
Approach: form views on interest rates, inflation, policy, and structural themes. Identify the sectors and names most exposed in each direction. Size positions by conviction and define what data would invalidate each view.
What I look for: clear macro or thematic drivers, identifiable winners and losers, and a way to monitor whether the thesis is playing out.
What I avoid: positions with no clear macro link, themes that are already fully priced in, and views I can't define a falsification test for.
Risk and sizing: size by conviction, always define the invalidation condition, and be willing to exit when the data turns against the thesis.
Research style: prioritize the strength of the macro driver, what's already priced in, the cleanest expressions of the theme, and the specific data to monitor. Frame everything in scenarios and probabilities, not single predictions.
```
## Step 3: Develop a thematic view
Use [Deep Research](/skills/deep-research) to build the top-down thesis:
```text theme={null}
Run Deep Research on [macro theme]. Explain the dynamics, the likely paths and their probabilities, what's already priced in, and which sectors and names are most exposed in each direction.
```
See [Macro research](/guides/research/macro-research) for the full workflow.
## Step 4: Find the cleanest expressions
```text theme={null}
For this theme, identify the cleanest ways to express the view in public markets, the names with the most direct exposure and the fewest offsetting factors, on both the long and short side.
```
## Step 5: Track the events that matter
Macro trading lives and dies by the calendar. Set up an [event tracker](/guides/scheduled/macro-event-tracker):
```text theme={null}
Each week, give me the key macro events coming up, what's expected, what's priced in, and how each could affect my themes and positions.
```
## Step 6: Monitor the thesis against the data
```text theme={null}
For each of my macro themes, track the incoming data and commentary. Flag anything that confirms the thesis, anything that challenges it, and tell me when the invalidation condition is approaching.
```
The discipline of macro trading is defining, in advance, what would tell you you are wrong, and then watching for it.
## Common mistakes
* **No falsification test.** A macro view you cannot disprove is a belief, not a tradeable thesis. Define what would make you exit.
* **Ignoring what's priced in.** The market often anticipates macro shifts. The opportunity is in the gap between consensus and reality, not the event itself.
* **Treating macro as certainty.** Macro is probabilistic. Size and frame positions accordingly, and avoid betting everything on a single forecast.
## Related
* [Macro research](/guides/research/macro-research) — the research workflow
* [Macro event tracker](/guides/scheduled/macro-event-tracker) — track the calendar
* [Deep Research](/skills/deep-research) — build the thesis
* [Build your Playbook](/guides/strategy/build-your-playbook) — the Playbook fundamentals
# Multi-Agent Strategies
Source: https://docs.driven.ai/guides/strategy/multi-agent-strategies
How to run multiple Driven Agents in parallel: separate value and momentum strategies, keep contexts clean, and compare how different approaches perform.
This guide covers running multiple Agents, each with its own strategy, context, and portfolio. Because each [Agent](/concepts/agents) is independent, you can run strategies that should not share assumptions side by side, and compare how they perform without one contaminating the other.
## Why use multiple Agents
Use separate Agents when standing preferences or research histories should stay distinct. Different portfolio rules alone do not require different Agents: one Agent can hold several paper portfolios, each with its own Playbook.
## Common multi-agent setups
* **Value Agent + momentum Agent** — two different philosophies, each with its own Playbook and rules
* **By market** — a US Agent, a Hong Kong Agent, an A-share Agent, each tuned to its market (note: paper trading covers US and Hong Kong only, so an A-share Agent is research-focused)
* **By horizon** — a long-term portfolio Agent and a short-term trading-experiment Agent
* **Research + monitoring** — one Agent for deep research, another dedicated to portfolio monitoring and alerts
## Step 1: Define each Agent's mandate
Give each Agent a distinct purpose and Profile. Create a paper portfolio when needed and give that portfolio its own Playbook. For a value Agent:
```text theme={null}
This agent runs a long-term value strategy. Save to its Profile: focus on profitable, durable businesses trading below intrinsic value, with strong balance sheets. Long holding periods. Avoid momentum-driven names.
```
For a separate momentum Agent:
```text theme={null}
This agent runs a momentum strategy. Save to its Profile: focus on names with strong relative strength and improving fundamentals. Shorter horizons. Defined invalidation levels on every position.
```
## Step 2: Keep contexts separate
The point of multiple Agents is isolation. Do not blur it — run value research in the value Agent and momentum research in the momentum Agent. Each builds its own history, portfolio, and intuition for its strategy.
## Step 3: Compare performance
Because each Agent has its own paper portfolio, you can compare approaches directly:
```text theme={null}
[In each agent] Summarize this strategy's paper portfolio performance and what's driving it.
```
Running the same period through two different strategies is one of the most useful things paper trading enables.
## Step 4: Let each Agent specialize
Keep each Agent's Profile and saved research current. Update the Playbooks of its portfolios when strategy rules change. More history does not by itself guarantee better analysis.
## Common mistakes
* **Too many Agents.** Every Agent is context to maintain. Create one per genuinely distinct strategy, not one per idea.
* **Blurring the mandates.** If you run momentum trades in your value Agent, you have lost the benefit of separation.
* **Not comparing.** The big payoff of multiple Agents is head-to-head comparison. Use it.
## When one Agent is enough
One Agent is enough when its strategies can share Agent-level preferences and research context. Use separate portfolios and Playbooks for different sizing, risk, or trading rules.
## Related
* [Agents concept](/concepts/agents) — how Agents work
* [Build your Playbook](/guides/strategy/build-your-playbook) — give each Agent its strategy
* [Value investing setup](/guides/strategy/value-investing-setup) — configure a value Agent
* [Growth investing setup](/guides/strategy/growth-investing-setup) — configure a growth Agent
# Value Investing Setup
Source: https://docs.driven.ai/guides/strategy/value-investing-setup
How to configure Driven for value investing: a Playbook, watchlist, screens, and scheduled tasks tuned for finding quality businesses trading below intrinsic value.
This guide is a worked example: configuring a Driven Agent end to end for a value investing strategy. It ties together the Playbook, screening, research, and scheduled tasks into one coherent setup you can adapt to your own approach.
For other styles, see [Growth investing setup](/guides/strategy/growth-investing-setup) and [Macro trading setup](/guides/strategy/macro-trading-setup).
## Step 1: Create a dedicated Agent
Value investing has its own assumptions and time horizon, so give it its own [Agent](/concepts/agents):
```text theme={null}
Create an agent named "Value Research" for a long-term value investing strategy.
```
## Step 2: Build the Playbook
The Playbook is what makes the Agent think like a value investor:
```text theme={null}
Add to this portfolio's Playbook:
Philosophy: Long-term value investing. Buy quality businesses trading below intrinsic value and hold for years.
Universe: US and Hong Kong listed companies. Profitable, established businesses.
What I look for: durable competitive advantages, consistent free cash flow, strong balance sheets, capable management, and a margin of safety between price and intrinsic value.
What I avoid: unprofitable companies, high leverage, businesses I can't understand, and hype-driven valuations.
Risk and sizing: maximum 10% per position, concentrated portfolio of high-conviction names, long holding periods.
Research style: prioritize business quality, moat durability, free cash flow, balance sheet strength, and valuation versus intrinsic value. Always state the bear case and what would break the thesis.
```
## Step 3: Set up value screening
```text theme={null}
Screen for US and Hong Kong companies with consistent free cash flow, low debt, returns on capital above their cost of capital, and valuations below their historical average and below peers. Rank by margin of safety.
```
This becomes a repeatable source of candidates. See [Screen stocks by fundamentals](/guides/screening/screen-stocks-by-fundamentals).
## Step 4: Research candidates the value way
For each candidate, run analysis focused on what matters to a value investor:
```text theme={null}
Analyze [TICKER] as a value investment. Cover the durability of the business, the moat, free cash flow generation, balance sheet strength, management quality, intrinsic value versus current price, and the bear case.
```
Then value it properly with [Valuation Matrix](/skills/valuation-matrix):
```text theme={null}
Run a Valuation Matrix on [TICKER]. Estimate intrinsic value with appropriate methods, cross-check against peers, and tell me the margin of safety at the current price.
```
## Step 5: Set up monitoring
Value investing is patient, so the monitoring is about theses, not daily moves:
```text theme={null}
Each month, review my value holdings. For each, check whether the business quality and thesis still hold, and flag any where the fundamentals have deteriorated. Ignore short-term price moves unless they create a new opportunity.
```
## Step 6: Build a watchlist of "wait for the price"
Value investors often find great businesses at the wrong price. Track them:
```text theme={null}
Add to my watchlist quality businesses I'd want to own at the right price, and alert me if any falls to an attractive valuation.
```
## Common mistakes
* **Confusing cheap with value.** A low multiple can signal a value trap. The Playbook emphasizes quality first, then price.
* **Impatience.** Value setups reward waiting. The monitoring is deliberately tuned to ignore noise.
* **Skipping the bear case.** Every value thesis should be stress-tested against what would make it wrong.
## Related
* [Build your Playbook](/guides/strategy/build-your-playbook) — the Playbook fundamentals
* [Valuation Matrix](/skills/valuation-matrix) — estimating intrinsic value
* [Screen stocks by fundamentals](/guides/screening/screen-stocks-by-fundamentals) — finding candidates
* [Multi-agent strategies](/guides/strategy/multi-agent-strategies) — run value alongside other strategies
# AI vs Traditional Research
Source: https://docs.driven.ai/learn/ai-vs-traditional-research
How AI-assisted investment research compares to traditional manual research: speed, breadth, consistency, and cost, and where human judgment still matters most.
Traditional investment research, done by hand or on a professional terminal, has been the standard for decades. AI-assisted research is changing the economics of it. This article compares the two honestly: where AI is a clear improvement, where traditional methods still win, and how the two combine.
## What traditional research looks like
Traditional research is a manual process: pull financials from filings or a data terminal, read the documents, build models in a spreadsheet, compare against peers by hand, and synthesize a view. It is thorough and gives the researcher deep familiarity with the company, but it is slow, and most of the time goes to gathering and organizing rather than thinking.
It also has a high skill floor. Reading filings, building models, and interpreting data well takes training that most individual investors do not have.
## What AI-assisted research changes
### Speed
The biggest change. Gathering data, organizing it, and producing a structured first read collapses from hours to minutes. You start from an organized analysis instead of a blank spreadsheet.
### Breadth
A person can deeply research a handful of names at a time. An AI system can screen a large universe, compare many companies, and synthesize many sources quickly, surfacing names and angles you would not have reached manually.
### Consistency
A human analyst's output varies with energy, time, and mood. A well-designed AI workflow applies the same framework every time, which makes analysis consistent and comparable across names. In Driven, this is what [Skills](/concepts/skills) provide.
### Cost and access
Professional research tools are priced for institutions. AI research platforms put institutional-style workflows within reach of individual investors at individual prices.
## Where traditional research still wins
AI does not make traditional research obsolete, and pretending otherwise is a mistake.
* **Deep, idiosyncratic judgment.** Some insights come from deep, sustained immersion in a company or industry, the kind a human builds over years. AI accelerates the gathering, but the deepest qualitative judgment is still human.
* **Original primary research.** Talking to customers, walking a store, reading between the lines of management's tone, this is human work.
* **Knowing what to ask.** AI answers the questions you pose. Knowing which questions matter is judgment, and it is yours.
* **Accountability.** A model is not responsible for your decisions. You are.
## The honest synthesis: combine them
The useful framing is not AI versus traditional, but AI plus judgment. Let AI do what it is good at, gathering, organizing, screening, synthesizing, producing a structured first read, and spend your time on what humans are good at: asking the right questions, exercising judgment, and making the decision.
In practice that looks like: use AI to screen a universe and produce first-pass analysis on the candidates, then apply your judgment to the shortlist, stress-test the theses, and decide. The AI compresses the grunt work; you keep the thinking.
## A caution on data quality
Financial feeds can contain delays, gaps, restatements, or normalization errors. AI can also misread correct data. Check the actual source, observation or reporting date, units, and original disclosure before relying on a figure.
## Where Driven fits
Driven is designed for the AI-plus-judgment workflow: live data, structured analysis, and automation that handle the gathering and synthesis, with output built to inform your decision rather than replace it. See [What is Driven](/get-started/what-is-driven).
## Related
* [How AI Agents work for investing](/learn/what-is-an-ai-investment-agent)
* [Building an investment workflow](/learn/building-an-investment-workflow)
# Building an Investment Workflow
Source: https://docs.driven.ai/learn/building-an-investment-workflow
How to build a repeatable investment research workflow with AI: define your strategy, set up screening, research candidates, manage a portfolio, and automate monitoring.
Most investors operate without a defined process, researching ad hoc, reacting to headlines, and making decisions inconsistently. A repeatable investment workflow fixes that, and AI makes a good workflow far easier to run. This article lays out how to build one, from defining your strategy to automating the recurring parts.
## Why a workflow matters
A workflow turns investing from a series of one-off reactions into a consistent process. It makes your decisions comparable (you evaluate every name the same way), it reduces the role of mood and impulse, and it frees your attention for judgment instead of logistics. The discipline is the edge.
AI lowers the cost of running a workflow, because the repetitive parts, gathering data, screening, monitoring, can be automated or compressed.
## The five stages of an investment workflow
### 1. Define your strategy
Everything starts here. What do you invest in, what do you look for, what do you avoid, how do you size positions and manage risk? Writing this down is the foundation; without it, every later step is arbitrary.
In Driven, use Profile for Agent-wide preferences and a [Playbook](/concepts/playbook) for the rules of a specific portfolio. See [Build your Playbook](/guides/strategy/build-your-playbook).
### 2. Source ideas through screening
A defined strategy implies a repeatable screen: the criteria that make a name a candidate for you. Running that screen regularly gives you a steady, consistent flow of ideas instead of whatever happened to catch your eye.
See [Screen stocks by fundamentals](/guides/screening/screen-stocks-by-fundamentals).
### 3. Research candidates consistently
Every candidate goes through the same analysis, so your evaluations are comparable. Cover the business, the financials, the valuation, and the risks, and always stress-test against the bear case. Consistency here is what makes your judgment reliable.
See [Research a stock](/guides/research/research-a-stock).
### 4. Build and manage a portfolio
Decisions become positions, and positions need management: sizing by your rules, tracking concentration and risk, and reviewing whether each holding's thesis still holds. A portfolio you do not review drifts.
See [Build a portfolio](/guides/portfolio/build-a-portfolio) and [Portfolio health check](/guides/portfolio/portfolio-health-check).
### 5. Automate monitoring
The recurring parts, the morning briefing, the watchlist monitor, the earnings calendar, the portfolio review, should run on their own. Automation is what lets a workflow persist without consuming all your time.
See [Create a market briefing](/guides/scheduled/create-a-market-briefing) and [Scheduled tasks](/concepts/scheduled-tasks).
## Putting it together
A complete workflow in practice:
* Your **Playbook** holds one portfolio's strategy rules; tasks run the recurring checks
* A **scheduled screen** feeds you candidates that fit it
* A **consistent research process** evaluates each candidate the same way
* A **paper or real portfolio** holds your decisions, managed by your rules
* **Scheduled monitors** keep you informed of what changes, without you watching
Each stage feeds the next, and the recurring parts run themselves. You spend your time on the judgment calls, not the logistics.
## Start small
You do not need to build all five stages at once. Start with a Playbook and one research workflow, add a scheduled briefing, then layer in screening and portfolio management as you go. A simple workflow you actually run beats an elaborate one you abandon.
## Where Driven fits
Driven is built to run this entire workflow in one place: strategy memory, screening, research, portfolio management, and automation, around live data and structured analysis. See [What is Driven](/get-started/what-is-driven) to start.
## Related
* [Build your Playbook](/guides/strategy/build-your-playbook)
* [How AI Agents work for investing](/learn/what-is-an-ai-investment-agent)
* [AI vs traditional research](/learn/ai-vs-traditional-research)
# Understanding Financial Data APIs
Source: https://docs.driven.ai/learn/understanding-financial-data-apis
A plain-language guide to financial data APIs: what they are, the categories of data they provide, why data quality varies, and why orchestration matters more than count.
Financial data APIs are the pipes that carry market and company data, quotes, statements, filings, holdings, into the tools investors use. You do not need to be technical to benefit from understanding them, because the quality and breadth of an AI research tool's data connections determine the quality of its answers. This article explains financial data APIs in plain language and what to look for.
## What a financial data API is
An API is a way for one piece of software to request data from another. A financial data API is a service that delivers financial information on request: ask for a company's revenue, its filings, or its current price, and the API returns it in a structured form a program can use.
When an AI research tool reports a company's margin, it is, behind the scenes, requesting that figure from a financial data API and relaying it to you. The figure is pulled, not remembered, which is exactly what you want.
## The main categories of financial data
* **Market data** — real-time and historical prices, volume, technical indicators
* **Fundamental data** — financial statements, ratios, earnings and dividend calendars
* **Filings and disclosures** — regulatory filings and company announcements
* **Ownership data** — institutional holdings, insider activity, fund flows
* **Specialized data** — beyond the basics, sources for signals like congressional trading, 13F institutional holdings, and Form 4 insider transactions in US markets; for A-shares, additional data such as Dragon & Tiger lists, margin trading, and share pledging
Different APIs specialize in different categories. No single source covers everything well.
## Why data quality varies
Not all financial data is equal, and the differences matter:
* **Timeliness** — some sources update in real time, others lag. A stale figure can mislead.
* **Accuracy and standards** — data quality varies by provider, and for cross-border companies, accounting-standard differences can make figures hard to compare unless the source aligns them.
* **Coverage** — a source strong in US equities may be thin on Hong Kong or A-shares.
* **Field definitions** — different sources define and label fields differently, which has to be reconciled.
This is why a tool's choice of sources, and how it handles their differences, affects the answers you get.
## Why orchestration matters more than count
A common claim is "we connect to X data sources." The count is less important than the orchestration. Connecting many sources is easy; making them work together is hard.
Source selection, comparable definitions, cross-checks, and missing-data handling all matter. Inspect whether those checks actually occurred; connecting or orchestrating APIs alone does not guarantee a reliable result.
When evaluating an AI research tool, ask not just how many sources it connects, but how it handles their differences.
## What this means for you
You will not call these APIs yourself, but understanding them helps you evaluate tools and read answers critically:
* Prefer tools that pull **live data** over those that recall from training
* Ask any tool to **cite its sources** so you can see what it used
* Be aware that **coverage varies by market**, confirm the markets you care about are well covered
* Value **orchestration** (cross-checking, gap-handling) over raw source count
## Where Driven fits
Driven connects to 260+ endpoints across 30+ categories and orchestrates them: each [Skill](/concepts/skills) knows which source to pull which field from, how to cross-check, and how to handle gaps. For Hong Kong and A-share stocks, statement data is sourced to match local accounting standards. See [Data sources](/reference/data-sources) and [Data and coverage](/concepts/data-and-coverage).
## Related
* [Data sources](/reference/data-sources)
* [Data and coverage](/concepts/data-and-coverage)
# What is an AI investment Agent
Source: https://docs.driven.ai/learn/what-is-an-ai-investment-agent
Understand how data, workflows, memory, and scheduled tasks support an investment research process.
An AI investment Agent uses models and tools to research markets, analyze evidence, retain context, and run configured tasks. Its usefulness depends on the data, workflows, and permissions available to it.
General AI products can also offer tools, memory, and scheduling. Compare the actual research environment and its configuration rather than assuming those features are exclusive to an investment Agent.
## Four parts to examine
| Part | What to check |
| - | - |
| Data | Markets, fields, sources, timestamps, reporting periods, and access rights |
| Workflows | Which methods and checks apply to your question, and how missing evidence is handled |
| Memory | What is saved, its scope, and how you can inspect or correct it |
| Scheduled work | Frequency, timezone, delivery, available tools, and execution permissions |
In Driven, [Skills](/concepts/skills) define reusable workflows. Profile holds Agent-wide preferences, Memory holds facts and notes, and a [Playbook](/concepts/playbook) holds one portfolio's strategy rules.
## A recurring workflow
Configure a pre-market briefing for a named watchlist, timezone, and delivery destination. The [scheduled task](/concepts/scheduled-tasks) can then check available data and produce a report while you are away. Inspect its run history, source dates, and missing information.
This work must be configured; creating an Agent alone does not establish every monitor, alert, or trading rule. Brokerage tools also require a supported connection and authorization. Review [permissions](/concepts/connectors#choose-a-permission-mode), especially for unattended work.
## Evaluate the result
Check whether the answer uses the correct listing, time window, currency, and data source. Distinguish retrieved facts, calculations, and model judgments. A professional feed can contain delays or errors, and a model can misinterpret it.
Model benchmarks such as [DrivenBench](/why-driven/drivenbench) can help compare tested workflows, but do not establish future task success or investment returns. You remain responsible for investment decisions.
## Related
* [What is Driven](/get-started/what-is-driven)
* [Building an investment workflow](/learn/building-an-investment-workflow)
* [Data and coverage](/concepts/data-and-coverage)
# Market Analysis Prompts
Source: https://docs.driven.ai/prompts/market-analysis-prompts
Copyable prompts for market, sector, and macro analysis in Driven: market pulse, sector scans, theme research, and connecting macro views to specific stocks.
Copyable prompts for reading the market, scanning sectors, and researching macro themes. The Agent selects data tools or relevant Skills such as [Market Pulse](/skills/market-pulse), [Sector Radar](/skills/sector-radar), and [Deep Research](/skills/deep-research) according to the task.
## Market pulse
```text theme={null}
Give me a market pulse for [US / Hong Kong / A-share] markets: index levels and moves, biggest movers, sector performance, and the main news driving today's action.
```
## What's moving and why
```text theme={null}
What's moving in the market this morning and why? Focus on the moves that actually matter, not just the biggest percentage changes.
```
## Sector scan
```text theme={null}
Use Sector Radar to compare native US sectors or industries. Show momentum, acceleration, valuation, the shared observation window, and unavailable dimensions.
```
## Sector rotation
```text theme={null}
Which sectors are seeing the strongest inflows and outflows recently, and is there a rotation pattern? What does it suggest about positioning?
```
## Sector deep dive
```text theme={null}
Do a deep dive on [sector]. What are the structural drivers, the key risks, the secular trends, and which public companies are best positioned for them?
```
## Theme research
```text theme={null}
Run Deep Research on [theme]. Identify the public companies exposed to it, the evidence supporting the thesis, the biggest risks, and which names have the cleanest exposure.
```
## Supply chain map
```text theme={null}
Map the [industry] supply chain and identify the public companies at each layer, then tell me where the value is accruing.
```
## Macro theme
```text theme={null}
Run Deep Research on [macro theme, e.g. sustained higher interest rates]. Explain the dynamics, the likely paths and probabilities, what's priced in, and which sectors and names are most exposed in each direction.
```
## Macro to specific names
```text theme={null}
Based on [macro theme], identify the public companies most positively and most negatively exposed, and explain the transmission channel for each.
```
## What's priced in
```text theme={null}
For [event or theme], what does the market already seem to be pricing in, and where's the gap between consensus and what could actually happen?
```
## Event preparation
```text theme={null}
[Major event, e.g. the next FOMC meeting] is coming up. Brief me on what's expected, what's priced in, the scenarios for the market reaction, and which of my holdings are most exposed.
```
## Challenge a market view
```text theme={null}
I think [your market view]. What's the strongest argument against it, and what data would tell me I'm wrong?
```
## Related
* [Macro research](/guides/research/macro-research) — the macro workflow
* [Sector deep dive](/guides/research/sector-deep-dive) — the sector workflow
* [Market Pulse Skill](/skills/market-pulse) and [Deep Research Skill](/skills/deep-research)
* [Scheduled task prompts](/prompts/scheduled-task-prompts) — automate market monitoring
# Driven prompt library for investment research
Source: https://docs.driven.ai/prompts/overview
A library of copyable prompts for investment research in Driven, organized by goal: stock research, screening, market analysis, scheduled tasks, paper trading, and Playbooks.
This is a library of copyable prompts for getting the most out of Driven, organized by what you are trying to do. Every prompt is a starting point: copy it, swap in your tickers and criteria, and adjust to fit. The Agent can use enabled [Skills](/skills/overview) or data tools according to the task.
## How to write a good prompt
The same habits that improve any prompt apply across all of these:
1. **Be specific about what you want.** "Analyze NVDA's data center margins and the risks to them" beats "tell me about NVDA." Specificity in, specificity out.
2. **Ask for evidence.** Request sources, assumptions, and flagged data gaps so you can audit the answer rather than taking it on faith.
3. **Name what matters to you.** Telling the Agent which drivers, metrics, or risks you care about focuses the analysis on your decision.
4. **Iterate in place.** The first answer is a draft. Push back: ask for the bear case, ask it to date claims, ask it to tighten scope.
5. **State your context.** Your horizon, your strategy, and your existing position all change the right answer. Say them.
## The collections
| Collection | For |
| - | - |
| [Stock research prompts](/prompts/stock-research-prompts) | Analyzing individual companies |
| [Screening prompts](/prompts/screening-prompts) | Finding stocks that match criteria |
| [Market analysis prompts](/prompts/market-analysis-prompts) | Reading the market, sectors, and macro themes |
| [Scheduled task prompts](/prompts/scheduled-task-prompts) | Automating briefings, monitors, and alerts |
| [Paper trading prompts](/prompts/paper-trading-prompts) | Testing ideas with simulated trades |
| [Playbook templates](/prompts/playbook-templates) | Strategy memory by investing style |
## A note on letting Driven route
Describe the objective, market, and desired output. You can name an enabled Skill or select it from the composer; the Agent may still hand off when another workflow better fits the task. See [Skill selection](/reference/skill-routing-logic).
## Related
* [Skills reference](/skills/overview) — what each Skill does
* [Guides](/guides/research/research-a-stock) — full walkthroughs that use these prompts
* [Skills concept](/concepts/skills) — how prompts map to Skills
# Driven paper trading prompts: Proposals, orders and cancellations
Source: https://docs.driven.ai/prompts/paper-trading-prompts
Copyable paper trading prompts for Driven: placing simulated trades, limit orders, position sizing, managing orders, building portfolios, and tracking performance.
Copyable prompts for testing ideas with paper trading, simulated trades on real market data. Replace `[TICKER]` and amounts to fit. These use your Agent's [portfolio and paper trading](/concepts/portfolio-and-paper-trading) context.
To review a proposal first, explicitly request no orders, then give a separate execution instruction. Simulated fills do not establish live fill results.
## Buy by percentage
```text theme={null}
Propose using 10% of my paper account cash to buy [TICKER]. Show the account, share quantity, and portfolio impact. Do not place an order.
```
## Buy by dollar amount
```text theme={null}
Propose a $5,000 purchase of [TICKER] in my USD paper account. Show quantity, portfolio impact, and risks. Do not place an order.
```
## Limit order
```text theme={null}
Place a limit order to buy [TICKER] at [price] in my paper account.
```
## Trim a position
```text theme={null}
Cut my [TICKER] paper position in half and explain the impact on portfolio concentration.
```
## Add to a position
```text theme={null}
Add to my [TICKER] position to bring it up to a [X]% weight in my paper portfolio.
```
## Cancel an order
```text theme={null}
List open [TICKER] orders and IDs in the specified paper portfolio. Let me select the orders, then cancel by order ID and check the results.
```
## Review open orders
```text theme={null}
Show me all my open paper orders and their current status.
```
## Check before trading
```text theme={null}
Before I buy [TICKER], show me how it would affect my portfolio's concentration, sector exposure, and risk.
```
## Build a starter portfolio
```text theme={null}
Build me a paper portfolio of [strategy, e.g. US large-cap quality companies]. Maximum 10% per position, at least 8 names, diversified across sectors. Show the rationale for each holding before building it.
```
## Track a thesis
```text theme={null}
Open a small paper position in [TICKER] to track this thesis, and remind me to review it in [timeframe] to see how it's playing out.
```
## Compare strategies
```text theme={null}
Summarize this agent's paper portfolio performance and what's driving it, so I can compare it against my other strategy.
```
## Rebalance
```text theme={null}
Propose a rebalance of my paper portfolio back toward my targets. Show the trades and how concentration and risk would change, then execute it once I confirm.
```
## Related
* [Place a paper trade](/guides/portfolio/place-a-paper-trade) — the full walkthrough
* [Build a portfolio](/guides/portfolio/build-a-portfolio) — construct one from scratch
* [Portfolio and Paper Trading concept](/concepts/portfolio-and-paper-trading) — how it works
* [Supported markets](/reference/supported-markets) — check trading coverage first
# Playbook Templates
Source: https://docs.driven.ai/prompts/playbook-templates
Copyable Playbook templates for Driven by investing style: value, growth, dividend income, macro, and active trading. Adapt one to make your Agent think like you.
Copyable [Playbook](/concepts/playbook) templates organized by investing style. Copy the one closest to your approach, then edit it to match how you actually invest. The Playbook is what makes your Agent apply your strategy automatically, so the more it reflects your real thinking, the better.
Name the portfolio when adding a template: "Update the Playbook for \[portfolio name]:". Put Agent-wide language, tone, and reporting preferences in Profile. These are strategy instructions, not enforced risk controls.
## Value investor
```text theme={null}
Philosophy: Long-term value investing. Buy quality businesses below intrinsic value and hold for years.
Universe: US and Hong Kong listed, profitable, established businesses.
What I look for: durable competitive advantages, consistent free cash flow, strong balance sheets, capable management, and a margin of safety.
What I avoid: unprofitable companies, high leverage, businesses I can't understand, hype-driven valuations.
Risk and sizing: max 10% per position, concentrated high-conviction portfolio, long holding periods.
Analysis criteria for this portfolio: prioritize business quality, moat durability, free cash flow, balance sheet strength, and valuation versus intrinsic value. Always state the bear case and what would break the thesis.
```
## Growth investor
```text theme={null}
Philosophy: Growth investing. Own durable, high-growth companies expanding into large markets, and hold through volatility while the growth thesis holds.
Universe: US and Hong Kong growth companies, primarily technology and software.
What I look for: high and durable revenue growth, large and expanding markets, improving unit economics, strong competitive positioning, a path to profitability.
What I avoid: decelerating growth without a reacceleration path, cash burn with no route to profit, growth dependent on a single fragile driver.
Risk and sizing: max 10% per position, tolerate volatility, exit if the growth thesis breaks.
Analysis criteria for this portfolio: prioritize growth durability, market size, unit economics, moats, and the path to free cash flow. Be explicit about valuation risk and what would break the growth thesis.
```
## Dividend / income investor
```text theme={null}
Philosophy: Income investing. Own stable businesses that pay and grow dividends, and hold for the income and compounding.
Universe: established, cash-generative companies with reliable dividends.
What I look for: a history of rising dividends, sustainable payout ratios, stable free cash flow, sound balance sheets, and defensible businesses.
What I avoid: stretched payout ratios, dividends funded by debt, businesses in structural decline, yield traps.
Risk and sizing: diversified across sectors, max 8% per position, prioritize dividend safety over headline yield.
Analysis criteria for this portfolio: prioritize dividend sustainability, free cash flow coverage, balance sheet strength, and business stability. Flag any risk to the dividend first.
```
## Macro / thematic trader
```text theme={null}
Philosophy: Top-down macro and thematic trading. Start from a macro or structural view, then find the cleanest expressions of it.
Approach: form views on rates, inflation, policy, and structural themes. Identify the most exposed sectors and names in each direction. Size by conviction and define what would invalidate each view.
What I look for: clear macro drivers, identifiable winners and losers, and a way to monitor whether the thesis is playing out.
What I avoid: positions with no clear macro link, fully priced-in themes, views I can't define a falsification test for.
Risk and sizing: size by conviction, always define the invalidation condition, exit when the data turns.
Analysis criteria for this portfolio: prioritize the strength of the driver, what's priced in, the cleanest expressions, and the data to monitor. Frame everything in scenarios and probabilities.
```
## Active trader
```text theme={null}
Philosophy: Active trading on shorter horizons, combining technical setups with fundamental and flow confirmation.
What I look for: strong relative strength, clean technical setups, confirming volume, and a fundamental or flow catalyst.
What I avoid: setups with no catalyst, thin-volume moves, and positions without a defined invalidation level.
Risk and sizing: defined risk on every position, hard invalidation levels, size by setup quality and risk.
Analysis criteria for this portfolio: prioritize the technical setup, the catalyst, volume and flow confirmation, and the precise level that invalidates the trade. Always state the risk before the reward.
```
## Building your own
If none of these fit, build from scratch. See [Build your Playbook](/guides/strategy/build-your-playbook) for a step-by-step guide. The essential elements are your universe, what you look for, what you avoid, your risk and sizing rules, and your research style.
## Related
* [Build your Playbook](/guides/strategy/build-your-playbook) — the full guide
* [Playbook concept](/concepts/playbook) — what the Playbook is
* [Value investing setup](/guides/strategy/value-investing-setup) and [Growth investing setup](/guides/strategy/growth-investing-setup) — worked examples
# Scheduled Task Prompts
Source: https://docs.driven.ai/prompts/scheduled-task-prompts
Copyable scheduled task prompts for Driven: pre-market briefings, watchlist monitors, earnings alerts, volume monitors, macro trackers, and portfolio reviews.
Copyable prompts for automating recurring work. Each is a complete scheduled task, swap in your tickers, times, and thresholds. Remember to anchor the time zone and define what's worth an alert. See [Scheduled Tasks](/concepts/scheduled-tasks) for the fundamentals.
Prompt conditions do not switch off notifications; results may still reach the configured channel. Test the schedule, time zone, and delivery behavior first.
## Daily pre-market briefing
```text theme={null}
Every weekday at 8 AM New York time, scan my watchlist and send a pre-market briefing. Include major news, price moves, earnings events, and risks to watch today. Deliver to Telegram.
```
## Short, scannable briefing
```text theme={null}
Every weekday at 7:30 AM New York time, give me a 5-bullet pre-market briefing on my watchlist: biggest move, key news, any earnings today, one risk to watch, and one opportunity. Deliver to Telegram.
```
## Market-wide overnight summary
```text theme={null}
Every trading day at 8 AM, summarize the overnight market action: index moves, notable sector moves, the main news driving them, and what to watch at the open.
```
## Watchlist monitor with thresholds
```text theme={null}
Monitor my watchlist during market hours. Alert me only when a name moves more than 5%, has significant news, or breaks a key technical level. Include what happened and why. If no event qualifies, report that briefly. Deliver to Telegram.
```
## Unusual volume monitor
```text theme={null}
During market hours, watch my watchlist for volume more than 2x its average. Flag each with the likely cause and whether it looks like buying or selling pressure. Deliver to Telegram.
```
## Earnings calendar
```text theme={null}
Each morning, tell me which of my watchlist names report earnings today or this week. For each, note the date, time (before or after market), what to watch, and what's already priced in. Deliver to Telegram.
```
## Post-earnings summaries
```text theme={null}
On any day a watchlist name reports, send me a summary after the release: results versus expectations, guidance, key metric trends, and whether the thesis still holds.
```
## Weekly macro calendar
```text theme={null}
Every Sunday evening, give me the week's key macro events with dates, expectations, and the main names or sectors exposed to each. Deliver to Telegram.
```
## Portfolio risk check
```text theme={null}
Each week, review my portfolio. Check concentration, sector and risk exposure, and flag any holding that has moved away from its thesis or crossed a risk threshold.
```
## Monthly portfolio health check
```text theme={null}
On the first trading day of each month, run a full portfolio health check: concentration, risk exposure, diversification, and any holdings that no longer fit my strategy.
```
## Conditional analysis
```text theme={null}
Every trading day, check my watchlist. If any name is down more than 8% from where I flagged it, run a quick analysis on what changed and whether the thesis still holds. If nothing crosses that line, report that no qualifying event was found.
```
## Multi-step Monday briefing
```text theme={null}
Every Monday morning: first, summarize how my portfolio performed last week. Then check whether any holdings hit risk thresholds. Then flag which holdings report earnings this week. Deliver as one combined briefing.
```
## Related
* [Scheduled Tasks concept](/concepts/scheduled-tasks) — how they work
* [Create a market briefing](/guides/scheduled/create-a-market-briefing) — the full walkthrough
* [Advanced scheduled tasks](/guides/scheduled/advanced-scheduled-tasks) — conditional and multi-step
* [Set up watchlist monitoring](/guides/scheduled/set-up-watchlist-monitoring) — threshold-based alerts
# Screening Prompts
Source: https://docs.driven.ai/prompts/screening-prompts
Copyable stock screening prompts for Driven: fundamental, growth, value, dividend, momentum, and capital-flow screens. Describe your criteria and get a ranked shortlist.
Copyable prompts for finding stocks that match your criteria. Adjust the markets, thresholds, and factors to fit your strategy. These route to the [Stock Screener](/skills/stock-screener).
## Quality and profitability
```text theme={null}
Find profitable [market] companies with revenue growth above 15%, positive free cash flow, gross margins above [X]%, and reasonable valuation. Rank the results and show the key metrics for each.
```
## Value screen
```text theme={null}
Screen [market] companies trading below their historical average valuation and below peers, with consistent free cash flow, low debt, and returns on capital above their cost of capital. Rank by margin of safety.
```
## Growth screen
```text theme={null}
Screen [market] companies with revenue growth above 20%, improving margins or a clear path to profitability, strong competitive positioning, and large addressable markets. Rank by the durability of the growth.
```
## Dividend growth
```text theme={null}
Find [market] dividend growers: a history of rising dividends, payout ratio under 60%, stable free cash flow, and a sound balance sheet.
```
## Momentum with quality
```text theme={null}
Find [market] stocks with strong relative strength trading near their highs with above-average volume, then keep only the ones that also have improving fundamentals and reasonable balance sheets.
```
## Improving fundamentals
```text theme={null}
Find [market] companies with improving margins and accelerating revenue over the last four quarters. Show the trend for each.
```
## Capital efficiency
```text theme={null}
Screen [market] companies with high and rising return on invested capital, strong free cash flow conversion, and disciplined capital allocation.
```
## Institutional inflows
```text theme={null}
Screen [market] companies seeing net institutional inflows over the past month, with improving fundamentals and reasonable valuation. Show the flow and the fundamental picture for each.
```
## Capital-flow multi-factor
```text theme={null}
Screen [market] companies with strong fundamentals, positive capital flow, and reasonable valuation. Include the flow and positioning context for the top names. (For A-shares, this can also draw on Dragon & Tiger and shareholder-structure data.)
```
## Turnaround candidates
```text theme={null}
Find [market] companies showing early signs of a fundamental turnaround: stabilizing or improving margins, returning revenue growth, and improving cash flow, after a period of weakness.
```
## Narrowing a broad screen
```text theme={null}
That returned too many names. Tighten it by adding [a constraint, e.g. return on invested capital above 15%] and re-rank.
```
## Loosening an empty screen
```text theme={null}
That returned nothing. Loosen the tightest constraint and show me what's closest to the criteria.
```
## Related
* [Screen stocks by fundamentals](/guides/screening/screen-stocks-by-fundamentals) — the full walkthrough
* [Find momentum plays](/guides/screening/find-momentum-plays) — technical screening
* [Stock Screener Skill](/skills/stock-screener) — what powers these prompts
* [Stock research prompts](/prompts/stock-research-prompts) — research the names you find
# Stock Research Prompts
Source: https://docs.driven.ai/prompts/stock-research-prompts
Copyable prompts for researching individual stocks in Driven: full analysis, valuation, bear cases, earnings prep, and investment memos. Swap in your ticker and go.
Copyable prompts for researching individual companies. Replace `[TICKER]` with your stock and adjust the focus to what matters for your decision. These route to [Stock Analysis](/skills/stock-analysis), [Valuation Matrix](/skills/valuation-matrix), and related Skills.
## Full company analysis
```text theme={null}
Analyze [TICKER]. Cover business overview, recent developments, revenue and margin trends, valuation, technical context, sentiment, key risks, and what would change the conclusion.
```
## Focused analysis on specific drivers
```text theme={null}
Analyze [TICKER], focused on [the drivers you care about, e.g. margin durability and competitive pressure]. Tell me what matters most for the stock over the next year and how to monitor it.
```
## The bear case
```text theme={null}
Make the strongest bear case for [TICKER]. What's the most compelling argument that this is a bad investment, and what evidence would confirm it?
```
## Bull and bear side by side
```text theme={null}
Give me the bull case and the bear case for [TICKER] side by side, then tell me which is better supported by the current evidence.
```
## Valuation
```text theme={null}
Run a Valuation Matrix on [TICKER]. Use multiple methods, cross-check against peers and management guidance, state the key assumptions, and flag data confidence. Tell me what growth and margin path the current price implies.
```
## Is it cheap or expensive
```text theme={null}
Is [TICKER] cheap or expensive right now? Compare its valuation to its own history and to peers, and explain what's driving the difference.
```
## Earnings preparation
```text theme={null}
[TICKER] reports earnings on [date]. Brief me: the key numbers and metrics to watch, what consensus expects, what guidance matters, what's already priced in, and what result would be bullish or bearish.
```
## Post-earnings read
```text theme={null}
[TICKER] just reported. Summarize results versus expectations, the guidance, how the key metrics trended, and whether the thesis still holds.
```
## Position review
```text theme={null}
Analyze [TICKER] as if I already own it at [price]. Should I add, hold, or trim, and what's the reasoning? What would change the answer?
```
## Thesis monitoring
```text theme={null}
What are the three things that matter most for [TICKER] over the next year, and what's the specific metric or event to monitor for each?
```
## Quality and durability check
```text theme={null}
Assess the business quality of [TICKER]: the competitive moat, the durability of revenue, the consistency of free cash flow, and the strength of the balance sheet.
```
## One-page memo
```text theme={null}
Turn your analysis of [TICKER] into a one-page investment memo: thesis, key evidence, valuation, risks, and what would invalidate it. Save it as a file.
```
## With sources
```text theme={null}
Analyze [TICKER] and cite the financial statement fields, filings, and market data used. Flag any data you could not retrieve rather than filling the gap.
```
## Related
* [Research a stock](/guides/research/research-a-stock) — the full walkthrough
* [Analyze before earnings](/guides/research/analyze-before-earnings) — the earnings workflow
* [Stock Analysis Skill](/skills/stock-analysis) — what powers these prompts
* [Screening prompts](/prompts/screening-prompts) — find candidates to research
# Driven financial data sources
Source: https://docs.driven.ai/reference/data-sources
Explore financial data categories for US, Hong Kong, and A-share research, including options, filings, ownership, macro, and source limitations.
This page groups the financial data used in Driven research. Actual coverage depends on the market, field, source, and entitlement. For boundaries, see [Supported markets](/reference/supported-markets).
## Why this matters
Ask the Agent to identify the source it actually used, the observation or reporting date, and missing data. A recent query does not establish that every returned figure is current or correct.
## Market data
* Real-time quotes
* Historical price data
* US and Hong Kong equity and ETF option expirations and chains
* Technical indicators
* Trading volume and liquidity data
## Fundamental data
* Financial statements (income statement, balance sheet, cash flow)
* Historical financials
* Earnings and dividend calendars
* Key ratios and metrics
For Hong Kong and A-share statements, check local reporting conventions, currency, accounting scope, and cumulative periods before comparing figures.
## Filings and disclosures
* SEC and exchange filings
* Company disclosures and announcements
## Ownership and positioning
* Institutional holdings and changes
* Insider activity
* ETF and fund data and flows
## A-share-specific data
Driven covers categories of A-share data that general tools usually miss:
**Capital flow**
* Institutional versus retail fund flow
* Dragon & Tiger (龙虎榜) lists
* Block trades (大宗交易)
* Margin trading (融资融券)
**Shareholder structure**
* Shareholder count (股东户数)
* Share capital breakdown (股本结构)
* Institutional holdings and changes
* Executive share changes (高管持股变动)
**Company intelligence**
* Thematic sector links (概念联动)
* Business segment commentary (主营点评)
* Share pledging (股权质押)
* Valuation bands (估值带)
* Historical earnings
* Peer comparison (同行对比)
## News and market intelligence
* Market news
* Community and social sentiment from Xueqiu and from X / Twitter (X content powered by Grok running xAI's official X search)
## How sources are used
Skills can define source selection, cross-checks, and missing-data handling. Review the actual evidence and execution rather than treating the workflow as a guarantee. See [Data and coverage](/concepts/data-and-coverage).
## Related
* [Data and coverage](/concepts/data-and-coverage) — the concept and best practices
* [Supported markets](/reference/supported-markets) — coverage boundaries
* [Insider Tracker](/skills/smart-money) — disclosed holdings and transaction research
# Driven FAQ: Data, Skills, trading and automation
Source: https://docs.driven.ai/reference/faq
Answers about Driven sign-in, Agents, Skills, options, paper and live trading, scheduled tasks, plans, and DrivenBench evaluations.
Common questions about Driven. For deeper detail, follow the links into the docs.
## What is Driven?
Driven is an AI Agent platform for investors. Your [Agent](/concepts/agents) combines financial data, specialized [Skills](/concepts/skills), research memory, portfolio strategies, and scheduled tasks in an investment workspace. See [What is Driven](/get-started/what-is-driven).
## How do I sign in to Driven?
You can sign in with Google or an email verification code. Email sign-in uses a six-digit code sent to your inbox and does not require a password. See [Quickstart](/get-started/quickstart#1-sign-in) for the steps.
## How is Driven different from ChatGPT or Claude?
General AI products can research the web, use memory, and run scheduled tasks. Driven packages financial data, investing Skills, Agent workspaces, portfolio strategy rules, and supported brokerage connections into a dedicated investment workflow. See [Driven vs general AI](/why-driven/driven-vs-chatgpt-claude-gemini).
## What is an Agent?
An Agent is a dedicated investment workspace with its own context, memory, research, scheduled tasks, and portfolios. You can run several Agents for different strategies. See [Agents](/concepts/agents).
## What is a Skill?
A Skill is a pre-built investing workflow with a defined framework, data sources, and quality checks. The Agent can choose an enabled workflow or you can select one; scope and available data still matter. See [Skills](/concepts/skills).
## What is a Playbook?
A Playbook is the investment strategy for a portfolio — its universe, watchlist, risk rules, position sizing, and screening. An Agent can run several portfolios, each with its own Playbook. See [Playbook](/concepts/playbook).
## What is a Profile?
An Agent's Profile is its identity and standing preferences — personality, tone, investing style, and how it works for you — saved and updated in chat. The Profile is who the Agent is; a Playbook is the strategy for a portfolio. See [Quickstart](/get-started/quickstart).
## What markets does Driven cover?
Broad research coverage, with emphasis on US and Hong Kong markets and A-share support as well. See [Supported markets](/reference/supported-markets).
## Where does the data come from?
From 260+ data endpoints across 30+ categories, pulled live rather than recalled from training. See [Data sources](/reference/data-sources). You can ask any Agent to cite the data it used.
## Is the trading real?
Driven supports two separate workflows. Built-in paper trading uses virtual funds. Live trading goes through a supported brokerage connector and requires the appropriate account, market, and order permissions. Check [connector permissions](/concepts/connectors#choose-a-permission-mode) before execution. See [Portfolio and paper trading](/concepts/portfolio-and-paper-trading) for simulation.
## Can Driven analyze options without a broker connection?
Yes. You can research supported US and Hong Kong options, compare contracts and strategies, and calculate payoff scenarios without connecting a broker. Real positions and trading history require a supported brokerage connector. Check quote timestamps and assumptions; the Options workspace may not be available to every account. See [Driven Options](/concepts/options).
## What does DrivenBench measure?
[DrivenBench](/why-driven/drivenbench) evaluates AI models on investment-agent tasks, including data handling, research, tool use, and execution. It reports capability scores, benchmark cost, latency, and separate regression checks. Results describe specific model versions and test conditions; they do not measure investment returns or guarantee success on your task.
## Can I use Driven outside the web app?
Yes. The same Agent is available on the web, mobile, Telegram, and WeChat. Quick questions work well on messaging channels; scheduled delivery is available through Telegram. Deep research and configuration are easier in the web app. See [Channels](/concepts/channels).
## Can I upload files?
Yes. Upload supported images, PDFs, audio, video, spreadsheets, `.docx` documents, and text or code files from the web app. Legacy `.doc` files are not supported. See [File uploads](/reference/file-uploads) for formats and limits.
## Can I run scheduled tasks?
Yes. Agents can run briefings, monitors, and alerts automatically on a schedule, delivered to the web app or Telegram. See [Scheduled tasks](/concepts/scheduled-tasks).
## How do I get better answers?
Be specific, ask for evidence and risks, name what matters to you, and iterate in the same conversation. See the [Prompt library](/prompts/overview).
## What plans are available?
Driven offers Free, Pro, Max, and Ultra. Compare current prices, workspace limits, and model access in [Plans and pricing](/get-started/plans-and-pricing). Scheduled tasks consume plan usage.
## Does Driven give financial advice?
Driven provides research and execution tools. Output is informational and does not constitute investment advice. Check evidence and risks; you remain responsible for investment decisions.
## Something isn't working
See [Troubleshooting](/reference/troubleshooting) for common issues, or contact support from inside the app.
## Related
* [What is Driven](/get-started/what-is-driven)
* [Quickstart](/get-started/quickstart)
* [File uploads](/reference/file-uploads)
* [Troubleshooting](/reference/troubleshooting)
# Upload files to a Driven Agent
Source: https://docs.driven.ai/reference/file-uploads
Upload supported images, PDFs, DOCX, spreadsheets, audio, video, and text to Driven. Check file-size limits and extraction limitations.
You can upload files to an Agent from the web app. Use the **Upload files** button in the chat input, or drag files into the composer when it is available.
Uploaded files become part of the conversation. The Agent can use them as context. Depending on the file type, file quality, selected model, and task, it may summarize content, extract data, or compare file content with live market data.
## Supported formats
Driven currently accepts these file types for chat uploads:
| Type | Supported formats | Common uses |
| - | - | - |
| Images | `.png`, `.jpg`, `.jpeg`, `.gif`, `.webp`, `.bmp`, `.tif`, `.tiff`, `.heic`, `.heif` | Screenshots, charts, broker statements, visual notes |
| PDF | `.pdf` | Reports, statements, exported documents |
| Word documents | `.docx` | Documents and text extraction |
| Audio | `.mp3`, `.wav`, `.ogg`, `.m4a`, `.flac`, `.aac`, `.opus` | Voice notes, recorded comments, short audio clips |
| Video | `.mp4`, `.webm`, `.mov`, `.avi`, `.mkv` | Screen recordings, walkthroughs, short clips |
| Spreadsheets and tabular files | `.xlsx`, `.xls`, `.csv`, `.tsv` | Excel workbooks, portfolio exports, tables |
| Markdown and text files | `.md`, `.markdown`, `.txt`, `.log` | Notes, prompts, exported text |
| Data and markup files | `.json`, `.yaml`, `.yml`, `.xml`, `.html`, `.htm`, `.svg` | Structured data, configuration, web exports |
| Code and config files | `.css`, `.js`, `.jsx`, `.ts`, `.tsx`, `.py`, `.go`, `.rs`, `.java`, `.c`, `.cpp`, `.h`, `.hpp`, `.sh`, `.bash`, `.zsh`, `.sql`, `.toml`, `.ini`, `.cfg`, `.conf`, `.env`, `.gitignore`, `Dockerfile`, `Makefile`, `.rb`, `.php`, `.swift`, `.kt`, `.scala`, `.r`, `.lua` | Source code, scripts, configuration |
The Agent's ability to interpret a file depends on the file quality, the selected model, and the task. For investment decisions, ask the Agent to cite what it read from the file and flag anything uncertain.
For extensionless files such as `Dockerfile` and `Makefile`, drag and drop may work better than the file picker if your browser does not show the file.
## Word documents
Driven accepts `.docx` files. Text extraction does not guarantee preservation of the original layout, images, or complex formatting.
Legacy `.doc` files are not supported. Convert them to `.docx`, or use one of these options:
* Export the Word document as `.pdf`, then upload the `.pdf` file.
* Copy the relevant text into the chat.
* Export tables as `.xlsx`, `.xls`, `.csv`, or `.tsv`.
* Save plain text as `.txt` or `.md`.
## Size limits
Current per-file limits are:
| File type | Limit |
| - | -: |
| Images | 20 MiB |
| Audio and video | 15 MiB |
| PDF | 50 MiB |
| Word documents, spreadsheets, tabular files, text, code, and config files | 20 MiB |
For large Excel workbooks, keep the workbook focused. If an upload fails because the workbook is too large or complex, export the relevant sheet as `.csv`.
Files saved to your workspace count toward workspace storage. See [Limits and quotas](/reference/limits-and-quotas).
## Tips for better results
* Upload the original text, `.csv`, `.tsv`, or spreadsheet file when you need precise tables or text.
* Upload `.xlsx` or `.xls` files when you want the Agent to read an Excel workbook.
* Use screenshots for visual errors, charts, or app states.
* Use supported video files for screen recordings and supported audio files for voice notes.
* Convert unsupported files to `.pdf`, `.xlsx`, `.xls`, `.txt`, `.md`, `.csv`, or `.tsv`.
* Ask the Agent to separate facts from assumptions when a file is incomplete.
* Keep investment decisions under your judgment. Uploaded files can inform analysis, but they do not replace your responsibility.
## Related
* [Channels](/concepts/channels) — use your Agent from Telegram and WeChat
* [Limits and quotas](/reference/limits-and-quotas) — storage and usage limits
* [Troubleshooting](/reference/troubleshooting) — common fixes
# Driven glossary: Agents, Skills and Playbooks
Source: https://docs.driven.ai/reference/glossary
Definitions of key Driven terms: Agent, Playbook, Skill, Scheduled Task, Paper Trading, Harness, and the market data concepts Driven covers.
Definitions of the terms used throughout Driven and this documentation.
## Core platform terms
**Agent**
A dedicated investment workspace with its own context, memory, research, scheduled tasks, and portfolios. Agents work independently of one another. See [Agents](/concepts/agents).
**Profile**
An Agent's identity and standing preferences — its personality, tone, investing style, and how it works for you. Save and update these preferences in chat; they guide work across the Agent. See [Quickstart](/get-started/quickstart).
**Memory**
Saved facts, research notes, preferences, and workspace references. It does not replace the Agent's Profile or a portfolio's Playbook.
**Playbook**
The investment strategy for a portfolio — its universe, watchlist, risk rules, position sizing, screening criteria, and buy/sell discipline. An Agent can run several portfolios, each with its own Playbook. See [Playbook](/concepts/playbook).
**Skill**
A purpose-built investing workflow with a defined analytical framework, data sources, steps, output format, and quality checks. The Agent can choose an enabled Skill according to the task, or you can select one explicitly; scope can require a handoff. See [Skills](/concepts/skills).
**Scheduled Task**
A prompt an Agent runs automatically on a schedule, for briefings, monitoring, and alerts, delivered to the web app or Telegram. See [Scheduled Tasks](/concepts/scheduled-tasks).
**Portfolio**
The cash, positions, orders, and performance tracked for an Agent. The Agent uses portfolio context when researching and monitoring risk. See [Portfolio and Paper Trading](/concepts/portfolio-and-paper-trading).
**Paper Trading**
Simulated trading with virtual funds on real market data. Lets you test ideas and practice without risking capital. See [Portfolio and Paper Trading](/concepts/portfolio-and-paper-trading).
## Conceptual terms
**Harness**
The runtime infrastructure around a model, data access, workflows, memory, and the ability to act, that turns a reasoning model into a useful Agent. Driven's framing: Agent = Model + Harness. See [Driven vs general models](/why-driven/driven-vs-chatgpt-claude-gemini).
**Agent = Model + Harness**
The idea that an AI Agent is a reasoning model plus everything around it that makes the reasoning useful. Models are broadly comparable; the harness is where the real difference lies.
## Built-in Skills
**Deep Research** — multi-source research with cited evidence and exportable reports. See [Deep Research](/skills/deep-research).
**Market Pulse** — macro, cross-asset, and event-transmission analysis. See [Market Pulse](/skills/market-pulse).
**Sector Radar** — US sector and industry rotation ranking. See [Sector Radar](/skills/sector-radar).
**Stock Screener** — multi-factor stock screening. See [Stock Screener](/skills/stock-screener).
**Stock Analysis** — balanced single-stock analysis. See [Stock Analysis](/skills/stock-analysis).
**Competitor Analysis** — peer comparison. See [Competitor Analysis](/skills/competitor-analysis).
**Valuation Matrix** — structured valuation with cross-checks. See [Valuation Matrix](/skills/valuation-matrix).
**Insider Tracker** — disclosed insider trades, institutional holdings, and ownership activity. See [Insider Tracker](/skills/smart-money).
**Portfolio Monitor** — holdings health, concentration, and risk. See [Portfolio Monitor](/skills/portfolio-monitor).
## Market data terms
Driven's coverage centers on US and Hong Kong markets, with A-share support as well. Some of the data it can pull:
**13F holdings** — quarterly disclosures of large institutional managers' US equity positions.
**Form 4 (insider transactions)** — insider buys and sells reported to the SEC.
**8-K filing** — a corporate disclosure of a material event.
**Earnings-call transcript** — the text of a company's earnings call.
**Congressional trades** — disclosed securities trades by US lawmakers.
**ETF reverse-lookup** — finding which ETFs hold a given stock.
**Analyst price target** — a sell-side analyst's projected price for a stock.
**Block trade** — a large transaction negotiated off the order book and reported separately.
A-share-specific data is covered too, including **capital flow (资金流向)**, the **Dragon & Tiger list (龙虎榜)** of most-active trading seats, and **margin trading (融资融券)**, leveraged buying and short selling.
## Related
* [What is Driven](/get-started/what-is-driven) — the overview
* [Skills reference](/skills/overview) — the Skills in detail
* [Data sources](/reference/data-sources) — the data behind the platform
# Driven usage, workspace limits and quotas
Source: https://docs.driven.ai/reference/limits-and-quotas
Understand how model access, Agent counts, portfolios, storage, and task usage are scoped.
Use [Plans and pricing](/get-started/plans-and-pricing) for the plan comparison and [current pricing](https://driven.ai/pricing) for published offers. Your account's available features and usage determine what you can run.
## How limits apply
| Limit | Scope |
| - | - |
| Agents | Active workspaces per user |
| Paper portfolios | Per Agent |
| Workspace storage | Per Agent, for uploaded and generated workspace files |
| Models | Plan, region, and current availability |
| Scheduled tasks | Each run consumes plan usage |
Model choice, task length, and tool work affect consumption. Displayed usage multipliers are plan comparisons, not a guaranteed number of questions or a direct conversion from benchmark dollars. [DrivenBench](/why-driven/drivenbench) measures a fixed evaluation suite.
## When an action is blocked
Check whether the issue is remaining usage, an Agent or portfolio limit, storage, model access, or a temporary service problem. For a scheduled task, inspect its run status. Narrow the task, reduce frequency, or upgrade when the limit is the cause.
Upload limits are separate from total storage. See [File uploads](/reference/file-uploads). External services can also impose their own permissions, quotas, or fees.
# How Driven selects and routes Skills
Source: https://docs.driven.ai/reference/skill-routing-logic
Help your Agent choose an appropriate workflow and understand when it may hand off to another Skill.
Describe your objective, market, and desired output. The Agent uses these details and the enabled Skills to choose a workflow. Selection is model-driven; a Skill is not a fixed keyword rule.
## Match the task to the scope
| Task | Relevant workflow |
| - | - |
| Analyze a known company or compare a few selected securities | [Stock Analysis](/skills/stock-analysis) |
| Estimate fair value, a price target, or implied growth | [Valuation Matrix](/skills/valuation-matrix) |
| Rank same-industry competitors by operating position | [Competitor Analysis](/skills/competitor-analysis) |
| Discover stocks matching criteria | [Stock Screener](/skills/stock-screener) |
| Investigate a theme or question across sources | [Deep Research](/skills/deep-research) |
| Examine insider trades or disclosed holdings | [Insider Tracker](/skills/smart-money) |
| Assess multi-holding portfolio risk | [Portfolio Monitor](/skills/portfolio-monitor) |
| Examine macro regimes or event transmission | [Market Pulse](/skills/market-pulse) |
| Compare US sector or industry strength | [Sector Radar](/skills/sector-radar) |
One request can use several workflows. A simple quote lookup may not need a Skill.
## Select a Skill explicitly
Name the Skill in your prompt, choose it from the composer, or find it through the `/` menu. See [Skills](/concepts/skills).
Explicit selection does not override the Skill's scope. For example, Deep Research can hand a pure fair-value request to Valuation Matrix. Sector Radar needs an explicit US market; a Hong Kong or A-share sector request uses a different workflow.
If the result does not match your goal, clarify the objective in the same conversation. Ask the Agent to state which workflow it used and why.
# Driven supported markets and trading coverage
Source: https://docs.driven.ai/reference/supported-markets
Check Driven research coverage for US, Hong Kong, and A-share markets, and distinguish data availability, built-in paper trading, and brokerage permissions.
Driven's deepest coverage — professional financial data and real-time quotes — is in the US, Hong Kong, and A-share (mainland China) markets, which have dedicated data sources. Beyond these, Driven can still research stocks in most major markets from public information and web sources, though data depth and timeliness vary with what's available. Paper trading is narrower than research: it currently supports the US and Hong Kong markets only. This page describes the coverage boundaries. Actual data availability depends on the source coverage for each market and asset type.
## Markets
Professional financial data and real-time quotes focus on three markets:
* **US** — US-listed stocks, ETFs, funds, and options on US-listed stocks or ETFs
* **Hong Kong** — Hong Kong listed equities and supported stock/ETF options
* **A-shares** — mainland China listed equities, with dedicated capital-flow, shareholder-structure, and company-intelligence data (see [Data sources](/reference/data-sources))
Stocks in most other major markets can still be researched from public sources, without the same data depth or real-time quotes.
## Research coverage
Research workflows can cover:
* Stocks
* ETFs
* Mutual funds
* Bonds
* US and Hong Kong equity and ETF options
* Forex
* Crypto
* Commodities
Coverage by asset type varies with source availability for each market.
## Paper trading coverage
Paper trading is narrower than research support. It currently covers the **US and Hong Kong** markets only — A-shares are not yet supported for paper trading, although research data is available subject to source coverage. You can often research a market or asset type that you cannot yet paper trade. Before building a workflow that depends on placing trades, confirm the current in-app account options for that market.
## Research coverage vs trading coverage
This distinction matters when building automated workflows. A [scheduled task](/concepts/scheduled-tasks) that researches a market will work even where paper trading is not yet supported; a task that places trades will not. Confirm both before relying on a workflow.
## Confirm current coverage
Coverage evolves. To check what is supported right now for a given market or asset type, ask the Agent directly:
```text theme={null}
Can you confirm whether [market / asset type] is supported for research and for paper trading? If not, explain the limitation.
```
## Related
* [Data sources](/reference/data-sources) — what data is available
* [Data and coverage](/concepts/data-and-coverage) — the concept
* [Portfolio and paper trading](/concepts/portfolio-and-paper-trading) — trading coverage context
# Troubleshoot Driven sign-in and workflows
Source: https://docs.driven.ai/reference/troubleshooting
Resolve Google or email-code sign-in issues, missing data, file uploads, scheduled tasks, and paper trading problems in Driven.
Solutions to common issues. If none of these resolve your problem, contact support from inside the app.
## Sign-in problems
* **No email code**: Check the email address and spam folder. After waiting for delivery, click **Try sending again** on the sign-in page.
* **Invalid or expired code**: Request a new code and use the latest email. To correct the address, click **Use a different email**.
* **Google sign-in fails inside another app**: Open Driven in your system browser and retry, or use email verification code sign-in.
See [Quickstart](/get-started/quickstart#1-sign-in) for both sign-in methods.
## The answer is too broad or generic
Usually a sign the prompt was too broad. Name what you care about, ask for specific drivers, metrics, and risks, and iterate in the same conversation rather than starting over:
```text theme={null}
Narrow this to [the specific drivers/metrics], and tell me what matters most for the decision.
```
See the [Prompt library](/prompts/overview) for sharper prompts.
## A financial figure looks wrong or missing
Ask the Agent to cite its sources and flag gaps. If it could not retrieve something, a good prompt makes it say so rather than guessing:
```text theme={null}
Cite the financial statement fields and data sources you used, and flag anything you could not retrieve.
```
If data for a market or asset type seems unavailable, confirm coverage, see [Supported markets](/reference/supported-markets).
## A Skill ran that I didn't expect
Driven routes to a Skill based on your prompt's intent. To get a different one, rephrase to signal the intent more clearly, or name the Skill explicitly:
```text theme={null}
Use [Skill name] instead, and [restate the task].
```
See [Skill routing logic](/reference/skill-routing-logic).
## A file upload failed
Confirm the file type and size limit. `.docx` is supported; legacy `.doc` is not. Convert `.doc` to `.docx` or PDF. Audio and video are limited to 15 MiB per file.
See [File uploads](/reference/file-uploads).
## A scheduled task didn't run or produced nothing
A few common causes:
* **No time zone** — an ambiguous time can misfire. Anchor it ("8 AM New York time").
* **Thresholds not met** — a run may find no qualifying events. Check its result; a prompt requesting silence does not guarantee suppression of channel notifications.
* **Scope too narrow or too broad** — review the task prompt and adjust.
Review the task's run logs to see what happened, and tighten the prompt if needed. See [Scheduled tasks](/concepts/scheduled-tasks).
## A paper limit order is stuck
Ask the Agent to list pending orders and confirm the portfolio and order IDs before canceling the intended orders:
```text theme={null}
Show me my open paper orders, then cancel the limit order on [TICKER].
```
See [Place a paper trade](/guides/portfolio/place-a-paper-trade).
## A paper trade was rejected
Paper trading coverage can be narrower than research coverage. Confirm the market and asset type are supported for trading before relying on a workflow:
```text theme={null}
Is [market / asset type] supported for paper trading?
```
See [Supported markets](/reference/supported-markets).
## The page or a browser extension is causing problems
If a browser translation extension interferes with the web app, try disabling it for the Driven site. If issues persist, contact support with your browser and extension details.
## A message went missing on Telegram or WeChat
Confirm the channel is connected, then review its conversation in the web app. Contact support if messages continue to go missing. Telegram and WeChat do not display connector approval prompts; tools can run automatically in these channels. See [Connector permissions](/concepts/connectors#choose-a-permission-mode).
## Related
* [FAQ](/reference/faq) — common questions
* [File uploads](/reference/file-uploads) — upload formats and limits
* [Prompt library](/prompts/overview) — getting better results
* [Scheduled tasks](/concepts/scheduled-tasks) — task configuration
* [Skill routing logic](/reference/skill-routing-logic) — how Skills are selected
# Competitor Analysis: Same-industry comparisons
Source: https://docs.driven.ai/skills/competitor-analysis
Driven's Competitor Analysis Skill compares a company against its peers across growth, margins, valuation, and positioning to show where it stands in its sector.
Competitor Analysis compares a company against its peers. Instead of analyzing one stock in a vacuum, it places the company side by side with its competitors so you can see where it actually stands on growth, profitability, valuation, and positioning.
## When to use it
* Deciding which name in a sector is the strongest
* Understanding a company's competitive position, not just its standalone numbers
* Sanity-checking a valuation against comparable companies
* Building the peer context for an investment thesis
## What it does
Competitor Analysis identifies the relevant peer set, pulls comparable metrics across the group, and lays out the comparison: relative growth, margins, valuation multiples, and qualitative positioning. The output shows not just the numbers but what the gaps mean.
## Prompt template
```text theme={null}
Compare [TICKER] against its main competitors. Cover relative growth, margins, valuation, market position, and which name looks strongest and why.
```
## Example
```text theme={null}
Compare NVDA against AMD and INTC. Who's winning on growth and margins, how do valuations compare, and what's the competitive trajectory?
```
## Tips
* **Let the Agent pick peers, then refine.** It will propose a peer set; if you want different comparables, name them.
* **Use it to break ties from a screen.** When [Stock Screener](/skills/stock-screener) returns several similar names, Competitor Analysis helps you choose.
* **Pair with valuation.** Follow up with [Valuation Matrix](/skills/valuation-matrix) on the name that looks strongest.
## Related
* [Compare two stocks](/guides/research/compare-two-stocks) — full walkthrough
* [Valuation Matrix](/skills/valuation-matrix) — structured valuation of the winner
* [Stock research prompts](/prompts/stock-research-prompts) — more templates
# Deep Research: Multi-source investment research
Source: https://docs.driven.ai/skills/deep-research
Driven's Deep Research Skill runs multi-source investment research across companies, sectors, and macro themes, cross-checks evidence, cites sources, and exports a report.
Deep Research is Driven's heaviest research workflow. It is built for questions that need evidence from many sources pulled together into a coherent, cited analysis, the kind of work that would take a human analyst hours of gathering and synthesis.
## When to use it
Reach for Deep Research when the question is broad or evidence-heavy:
* Company deep dives
* Industry maps and competitive landscapes
* Earnings-season summaries across multiple names
* Policy or macro event impact analysis
* Theme research ("who benefits from X")
For a single metric, ask for a direct data lookup. Use [Stock Analysis](/skills/stock-analysis) for a focused security question, or [Market Pulse](/skills/market-pulse) for macro and event transmission.
## What it does
Deep Research defines the scope of the question, collects evidence from multiple data sources, separates facts from opinion, cross-checks key figures, summarizes the risks, and produces an analyst-style brief. Because it runs as a multi-step workflow, important numbers are verified before they appear in the final output, which reduces calculation, unit-conversion, and source errors.
Ask the final report to retain the key evidence, assumptions, and unresolved gaps. Review its citations and supporting material.
## Prompt template
```text theme={null}
Run Deep Research on [topic]. Define the scope, collect evidence from multiple sources, separate facts from opinions, cite sources, summarize risks, and produce an analyst-style brief.
```
## Example
```text theme={null}
Run Deep Research on the AI data center power bottleneck. Identify public companies exposed to the theme, what evidence supports the thesis, and the biggest risks.
```
## Export the report
When the research is long, ask for a downloadable file:
```text theme={null}
Save the final report as a file in my workspace.
```
The report lands in the Agent's files, where you can revisit or share it.
## Tips
* **Set the scope explicitly.** Naming what to include and exclude keeps a broad topic from sprawling.
* **Ask for the risks and the disconfirming evidence**, not just the thesis. The most useful Deep Research output tells you what would make the idea wrong.
* **Chain it.** Run Deep Research to map a theme, then use [Stock Analysis](/skills/stock-analysis) or [Valuation Matrix](/skills/valuation-matrix) on the strongest names it surfaces.
## Related
* [Run deep research](/guides/research/run-deep-research) — full walkthrough
* [Stock research prompts](/prompts/stock-research-prompts) — more templates
* [Market analysis prompts](/prompts/market-analysis-prompts) — theme and macro prompts
# Market Pulse: Macro and event analysis
Source: https://docs.driven.ai/skills/market-pulse
Analyze macro regimes, cross-asset changes, broad-market health, and event transmission using dated evidence.
Market Pulse investigates macro conditions and how events transmit across markets. It first checks the premise, geography, and observation window, then uses comparable growth, inflation, rates, and market evidence.
## When to use it
* Examine a macro regime or cross-asset divergence
* Assess how CPI, employment, rates, tariffs, or policy affect assets
* Investigate a broad-market index or ETF move
A macro question can concern one security. An issuer-specific move belongs to [Stock Analysis](/skills/stock-analysis); US sector ranking belongs to [Sector Radar](/skills/sector-radar). Routine quote lookups or news briefings do not require Market Pulse.
## Example
```text theme={null}
Use Market Pulse to examine how the latest US inflation release affected
Treasury yields, the dollar, and equities. Compare the same event window,
verify the release date, and separate observations from causal hypotheses.
```
## What to inspect
The result should state dated evidence, units, a common comparison window, and missing or conflicting signals. A macro regime or causal explanation is a working hypothesis, not proof from a fixed asset-move pattern. Ask what would confirm or weaken it.
Use [Deep Research](/skills/deep-research) for broader source synthesis and [scheduled tasks](/concepts/scheduled-tasks) for recurring checks. Investment decisions remain yours.
# Driven investing Skills reference
Source: https://docs.driven.ai/skills/overview
Choose current Driven built-in Skills by task and understand their scope.
The public built-in Skills are listed below. Availability depends on account access and what is enabled for the current Agent. See [Options](/concepts/options) for position sync and weekly review.
| Skill | Use it for |
| - | - |
| [Deep Research](/skills/deep-research) | Multi-source research |
| [Market Pulse](/skills/market-pulse) | Market and macro context |
| [Sector Radar](/skills/sector-radar) | US sector and industry rotation |
| [Stock Screener](/skills/stock-screener) | Stock discovery with stated criteria |
| [Stock Analysis](/skills/stock-analysis) | Known-security analysis and small comparisons |
| [Competitor Analysis](/skills/competitor-analysis) | Same-industry competitive ranking |
| [Valuation Matrix](/skills/valuation-matrix) | Fair value and implied expectations |
| [Insider Tracker](/skills/smart-money) | Disclosed transactions and ownership |
| [Portfolio Monitor](/skills/portfolio-monitor) | Multi-holding risk diagnostics |
| [0DTE Option Opportunities](/concepts/options#research-contracts-and-opportunities) | US same-day options research |
| [Unusual Option Activities](/concepts/options#research-contracts-and-opportunities) | Observable US contract activity |
| [Earnings Option Opportunities](/concepts/options#research-contracts-and-opportunities) | Earnings options in one selected market |
| [Quant Calculator](https://driven.ai/skills/data-lab) | Financial calculations and tabular data |
## Select and use
Describe the task or select an enabled Skill. Scope matters: a request can combine workflows, and a named Skill may hand off to another. See [Skills](/concepts/skills) and [Skill routing](/reference/skill-routing-logic).
Check sources, timestamps, assumptions, and missing evidence. Install approved community Skills through [Skill Store](/concepts/skill-store); installation applies to the Agent you choose.
# Portfolio Monitor: Holdings and risk analysis
Source: https://docs.driven.ai/skills/portfolio-monitor
Diagnose long-only portfolio concentration, correlation, and historical risk contributions with Driven; distinguish diagnostics from optimization or trading.
Portfolio Monitor checks the health of your holdings. It looks across your portfolio, not at one stock, and tells you where the risks are: concentration, exposure, names that have drifted from your thesis, and anything that needs attention.
Current diagnostics cover multi-holding, long-only portfolios. Analyze supplied holdings or a stated target-weight scenario; this workflow does not optimize weights, generate an efficient frontier, or execute trades.
## Framework
Portfolio Monitor is grounded in **Markowitz mean-variance theory**: it runs a covariance-based risk decomposition to reveal each holding's estimated contribution to historical portfolio risk. That contribution is driven by volatility and correlation, not by position weight alone, so a small holding correlated with everything else can carry more risk than its size suggests. Because this runs as real code, the risk attribution is reproducible rather than a free-form judgment.
## When to use it
* Reviewing whether your portfolio is too concentrated
* Understanding your risk and sector exposure
* Finding holdings that no longer fit your strategy
* Doing a periodic health check on positions
## What it does
Portfolio Monitor pulls your current positions and analyzes them as a whole: concentration by position and sector, risk exposure, diversification, and how each holding stands relative to your strategy. It uses the portfolio context attached to your Agent, and it can weigh holdings against the rules in your [Playbook](/concepts/playbook).
## Prompt template
```text theme={null}
Review my portfolio. Check concentration, sector and risk exposure, diversification, and flag any holdings that need attention or no longer fit my strategy.
```
## Example
```text theme={null}
Do a health check on my portfolio. Am I too concentrated anywhere? What are my biggest risks, and which positions should I reconsider?
```
## Tips
* **Keep your Playbook current.** Portfolio Monitor is sharper when it knows your risk rules and position-sizing limits; those live in the [Playbook](/concepts/playbook).
* **Run it on a schedule.** A monthly or weekly portfolio review makes a good scheduled task. See [Scheduled tasks](/concepts/scheduled-tasks) for configuration.
* **Act on the concentration flags.** The most common finding is hidden concentration, several positions exposed to the same driver. Ask the Agent to map that overlap.
## Related
* [Portfolio health check](/guides/portfolio/portfolio-health-check) — full walkthrough
* [Risk exposure analysis](/guides/portfolio/risk-exposure-analysis) — deeper on risk
* [Rebalancing strategies](/guides/portfolio/rebalancing-strategies) — acting on the review
# Sector Radar: US sector and industry rotation
Source: https://docs.driven.ai/skills/sector-radar
Compare native US sectors or industries using momentum, acceleration, and valuation evidence.
Sector Radar ranks **US sectors or industries**, helping you decide where to focus before choosing individual stocks. It compares entities on a common data basis using momentum, acceleration, and available valuation evidence.
## Scope
* Specify the US market. If the market is unclear, the Agent asks before fetching data.
* For Hong Kong or A-share sector research, use Market Pulse for context or Stock Screener for a named industry.
* For companies within an industry, use Stock Screener or Competitor Analysis.
* For named ETFs or funds, use Stock Analysis rather than treating their returns as sector returns.
## Example
```text theme={null}
Use Sector Radar to compare US sectors. Show momentum, acceleration,
valuation, the observation window, and any unavailable dimensions.
```
The calculation uses comparable sector or industry data. Its compounded constituent-average performance is a synthetic measure, not an official index or investable ETF return. Missing comparable data can remove a scoring dimension or prevent a ranking.
Treat the result as research context, not a trade instruction. Recheck the data and your risk constraints before making an allocation decision.
## Related
* [Stock Screener](/skills/stock-screener) — discover individual candidates
* [Competitor Analysis](/skills/competitor-analysis) — compare operating competitors
* [Market Pulse](/skills/market-pulse) — broad market and macro context
# Insider Tracker: Disclosed trades and holdings
Source: https://docs.driven.ai/skills/smart-money
Inspect disclosed insider trades, institutional holdings, congressional trades, and market-specific ownership activity.
Insider Tracker examines **disclosed transactions and holdings**. Each source has its own reporting period, population, and publication delay. Use it to investigate a specific signal rather than infer a universal view of what informed investors are doing.
## What you can research
* US Form 4 insider transactions, separating voluntary purchases and sales from awards, exercises, gifts, and tax-related events
* US 13F holdings and changes for a disclosed period
* Disclosed US congressional trades
* Supported Hong Kong shareholder records
* A-share Dragon & Tiger disclosures and the scope of institutional-seat activity
The Skill uses the route relevant to your request. These disclosures do not establish current positions or the intent behind every transaction. Generic capital-flow analysis may use separate data tools.
## Example
```text theme={null}
Use Insider Tracker to review NVDA insider transactions.
Separate voluntary purchases and sales from administrative events.
Show transaction dates, filing dates, source coverage, and missing data.
```
For a 13F comparison, specify the filing period and ask whether the disclosed population is complete. Do not combine differently timed signals into a single buying or selling verdict without stating the limits.
## Related
* [Track institutional holdings](/guides/screening/track-institutional-holdings)
* [Insider trading signals](/guides/screening/insider-trading-signals)
* [Data and coverage](/concepts/data-and-coverage)
# Stock Analysis: Company and security research
Source: https://docs.driven.ai/skills/stock-analysis
Analyze a company or a few selected securities with Driven. Match financial, valuation, technical, and risk methods to the question and available evidence.
Stock Analysis produces a balanced, structured view of a single company. It is the workhorse Skill for "should I understand this stock better?", covering the business, the financials, the valuation, and the risks in one coherent answer.
## When to use it
* Getting a complete read on one company
* Forming or pressure-testing a thesis on a stock
* Preparing before earnings or before a decision
For a broad comparison of two or three selected securities, use Stock Analysis. For same-industry competitive ranking, use [Competitor Analysis](/skills/competitor-analysis). For a valuation-focused deep dive, use [Valuation Matrix](/skills/valuation-matrix). For broad theme research, use [Deep Research](/skills/deep-research).
## Framework
Stock Analysis chooses the smallest workflow that answers your question. A full company analysis can cover business quality, financial health, valuation context, price trends, and catalysts. A technical question may need only adjusted prices and indicators. Banks, insurers, REITs, and funds require different methods. It does not require every request to run a fixed financial-score model.
## What it does
Stock Analysis pulls the company's fundamentals, recent developments, revenue and margin trends, valuation, technical context, and sentiment, then organizes them into a structured analysis that separates facts from assumptions and lays out the key risks and what would change the conclusion.
## Prompt template
```text theme={null}
Analyze [TICKER]. Cover business overview, recent developments, revenue and margin trends, valuation, technical context, sentiment, key risks, and what would change the conclusion.
```
## Example
```text theme={null}
Analyze MSFT. Focus on Azure growth, AI capex, margin durability, valuation, and risks from cloud competition.
```
## Review checklist
A good Stock Analysis answer should pass these checks:
* Does it separate facts from assumptions?
* Are recent events dated?
* Are financial figures sourced?
* Are the risks concrete, not generic?
* Does the conclusion say what evidence would change it?
If any check fails, push back in the same conversation.
## Follow-up prompts
```text theme={null}
Compare this thesis against the bear case.
```
```text theme={null}
Turn this into a one-page investment memo.
```
## Tips
* **Focus the analysis.** Naming what you care about ("focus on margins and competition") produces a sharper read than a generic "analyze this."
* **Ask for the disconfirming case.** The bear case is often more valuable than the bull case you already believe.
## Related
* [Research a stock](/guides/research/research-a-stock) — full walkthrough
* [Analyze before earnings](/guides/research/analyze-before-earnings) — the earnings-specific version
* [Stock research prompts](/prompts/stock-research-prompts) — more templates
# Stock Screener: Find investment candidates
Source: https://docs.driven.ai/skills/stock-screener
Screen US, Hong Kong, and A-share stocks against stated criteria using available data. Review coverage, observation windows, and missing fields.
Stock Screener finds stocks that match the criteria you describe. Instead of clicking through filter menus, you tell the Agent in plain language what you are looking for and it returns the names that fit, with the metrics that matter.
## When to use it
* Finding investment candidates that meet specific criteria
* Narrowing a large universe down to a workable shortlist
* Combining fundamental, technical, and flow factors in one screen
* Discovering names you would not have thought to look at
## What it does
Stock Screener translates your description into a multi-factor filter, applies it against live data, and returns the matching companies ranked and annotated with the relevant metrics. Because it draws on Driven's full data coverage, you can screen on fundamentals, technicals, and, for A-shares, capital-flow and shareholder-structure factors in a single request.
## Prompt template
```text theme={null}
Screen [market] companies with [fundamental criteria], [technical criteria], and [valuation criteria]. Rank the results and show the key metrics for each.
```
## Examples
```text theme={null}
Find profitable US software companies with revenue growth above 15%, positive free cash flow, and reasonable valuation.
```
```text theme={null}
Screen A-share companies seeing net institutional inflows over the past month with improving margins and reasonable valuation.
```
## Tips
* **Combine factor types.** The screener's strength is mixing fundamentals, technicals, and flow in one screen. Use it.
* **Start broad, then tighten.** Run a loose screen, see the shape of the results, then add constraints to narrow it.
* **Hand results to another Skill.** Take the shortlist into [Stock Analysis](/skills/stock-analysis) or [Valuation Matrix](/skills/valuation-matrix) to go deep on the best candidates.
* **Watch for over-filtering.** If a screen returns nothing, loosen the tightest constraint first.
## Related
* [Screen stocks by fundamentals](/guides/screening/screen-stocks-by-fundamentals) — full walkthrough
* [Find momentum plays](/guides/screening/find-momentum-plays) — technical-factor screening
* [Screening prompts](/prompts/screening-prompts) — more templates
# Valuation Matrix: Fair value and assumptions
Source: https://docs.driven.ai/skills/valuation-matrix
Research company fair value and implied expectations with Driven. Inspect valuation assumptions, comparable evidence, and instrument-specific limitations.
**Valuation Matrix** researches fair value and implied expectations for one listed company. It selects methods suited to the company and available evidence, states assumptions, and reports missing inputs. Peer and guidance checks depend on comparable data.
## Framework
Valuation Matrix is grounded in **DCF and reverse DCF**: it discounts projected cash flows to an intrinsic value, then inverts the model to back out the growth and margin assumptions the current price already embeds. It cross-checks that output against **Damodaran valuation multiples** so the result is anchored to peer-implied benchmarks rather than free-form guesswork.
## When to use it
* Putting a defensible value on a company
* Pressure-testing whether a stock is cheap or expensive
* Building the valuation section of an investment thesis
* Comparing a company's valuation against its own history and its peers
## What it does
Valuation Matrix pulls the financial data, applies relevant valuation methods, and, importantly, runs the cross-checks a careful analyst would: it compares the output against peer valuations, checks it against management guidance, and notes the confidence level of the underlying data. Inspect these checks and assumptions before using the estimate.
Peer comparisons and guidance checks depend on available, comparable evidence. Ask the Agent to identify checks it completed and gaps it could not resolve. Banks, insurers, REITs, ETFs, and funds may need a different framework; do not force a generic company DCF onto them.
## Prompt template
```text theme={null}
Run a Valuation Matrix on [TICKER]. Use multiple valuation methods, cross-check against peers and management guidance, state the key assumptions, and flag data confidence.
```
## Example
```text theme={null}
Value XIAOMI (1810.HK). Use appropriate methods for its business mix, cross-check against peers, be explicit about assumptions, and flag where the data is uncertain.
```
## Tips
* **Check the assumptions first.** A valuation is only as good as its inputs; ask the Agent to surface the key assumptions so you can challenge them.
* **Mind accounting standards and units.** For cross-market names, confirm the analysis is not mixing accounting standards or mismatching reporting periods. Ask the Agent to state which standard and period it used.
* **Re-run with your own assumptions.** If you disagree with a growth or margin assumption, give the Agent yours and have it re-run.
## Related
* [Research a stock](/guides/research/research-a-stock) — the broader single-stock workflow
* [Competitor Analysis](/skills/competitor-analysis) — peer context for the valuation
* [Stock research prompts](/prompts/stock-research-prompts) — more templates
* [Analyze before earnings](/guides/research/analyze-before-earnings) — apply the valuation ahead of an earnings report
# Driven vs AlphaSense
Source: https://docs.driven.ai/why-driven/driven-vs-alphasense
Compare AlphaSense intelligence and research agents with Driven personal investment workspaces, portfolio context, Skills, and recurring reviews.
AlphaSense combines market intelligence, professional content, and internal knowledge with search, deep research, and Workflow Agents. Custom workflows can also run on a schedule. Driven organizes personal investment goals, portfolio rules, and ongoing research within Agent workspaces.
## Compare your workflow needs
| Dimension | AlphaSense | Driven |
| - | - | - |
| Content and organization | Entitled research content and institutional knowledge | Built-in financial data, user materials, and Agent records |
| Reusable work | Research Workflow Agents and scheduled workflows | Investing Skills and Agent scheduled tasks |
| Portfolio context | Check actual product capabilities and entitlements | Store a paper portfolio's Playbook and review its positions |
## A practical workflow
Both products can help organize earnings evidence and produce research. In Driven, continue by checking the conclusion against the selected portfolio's rules and setting the next review. AlphaSense's proprietary content remains subject to its licensing; a Driven subscription does not include those rights.
## Requirements and fit
Evaluate AlphaSense for professional content libraries, internal knowledge, and organizational research workflows. Evaluate Driven for ongoing work around personal goals, positions, and strategies. Compare content and work organization rather than assuming only one product has AI or automation.
Features, regions, subscriptions, and data rights can change; verify your current account. Live trading needs supported broker tools and authorization. Scheduled and messaging-channel workflows can run connector tools automatically; review [permissions](/concepts/connectors#choose-a-permission-mode). Research output is informational, and investment decisions remain yours.
## Related
* [Full comparison on Driven](https://driven.ai/comparisons/alphasense-vs-driven)
* [What is Driven](/get-started/what-is-driven)
* [Data and coverage](/concepts/data-and-coverage)
* [Official product information](https://www.alpha-sense.com/platform/)
# Driven vs Bloomberg Terminal
Source: https://docs.driven.ai/why-driven/driven-vs-bloomberg-terminal
Compare Bloomberg Terminal data, research, and professional tools with Driven investment Agents, Skills, portfolio rules, and broker connections.
Bloomberg Terminal combines professional market data, news, research, analytics, and communications, and also offers AI research capabilities. Driven focuses on personal investment workflows through Agent workspaces, Skills, strategy rules, and supported broker connections.
## Compare your workflow needs
| Dimension | Bloomberg Terminal | Driven |
| - | - | - |
| Focus | Professional data, content, communication, and institutional workflows | Personal research, strategy records, position analysis, and recurring checks |
| Research environment | Use entitled Terminal functions and content | Agent conversations, investing Skills, and files |
| Execution | Depends on licensed services and permissions | Built-in paper accounts; live orders through supported broker connectors |
## A practical workflow
If you have Bloomberg access, continue using its entitled research and data tools. To track a thesis in Driven, supply material you are authorized to use, define portfolio rules, and schedule reviews. Do not assume Terminal content can be read through an existing connector or redistributed.
## Requirements and fit
Evaluate Bloomberg against your professional communication, content, and institutional-tool requirements. Evaluate Driven for a self-service investment workspace and recurring research. Compare entitlements, data scope, and workflows rather than treating the products as interchangeable on price.
Features, regions, subscriptions, and data rights can change; verify your current account. Live trading needs supported broker tools and authorization. Scheduled and messaging-channel workflows can run connector tools automatically; review [permissions](/concepts/connectors#choose-a-permission-mode). Research output is informational, and investment decisions remain yours.
## Related
* [Full comparison on Driven](https://driven.ai/comparisons/bloomberg-terminal-vs-driven)
* [What is Driven](/get-started/what-is-driven)
* [Data and coverage](/concepts/data-and-coverage)
* [Official product information](https://professional.bloomberg.com/products/bloomberg-terminal/access/bloomberg-professional-app/)
# Driven vs ChatGPT, Claude, and Gemini
Source: https://docs.driven.ai/why-driven/driven-vs-chatgpt-claude-gemini
Compare a dedicated investment workspace with general AI research tools by data, workflows, context, and execution requirements.
General AI products can support investment research, file analysis, coding, and connected tools. ChatGPT also offers [research and memory](https://help.openai.com/en/articles/8590148-memory-faq) and [scheduled tasks](https://help.openai.com/en/articles/10291617-scheduled-tasks-in-chatgpt). Features vary by product, plan, region, and configuration.
Driven brings financial data, investing Skills, Agent context, and account workflows together for recurring investment research. Choose based on the process you need and the integrations you want to maintain.
## What is configured for investment work
| Dimension | General AI tools | Driven |
| - | - | - |
| Data | Web, files, and connected services; access depends on configuration | Built-in US, Hong Kong, and A-share research data, including US/HK options; sources and entitlements still vary |
| Methods | Flexible research, code, and reusable instructions | Investing Skills for valuation, screening, disclosed activity, portfolio risk, and options |
| Context | Memory, projects, files, and other product-specific context features | Agent Profile and memory, plus a Playbook for each paper portfolio |
| Recurring work | Scheduled or background work where supported | Scheduled tasks using the Agent's available context and tools |
| Accounts | Depends on added integrations | Built-in paper trading and supported brokerage connectors for live positions and orders |
## When Driven fits
Use Driven when you want an investment workspace with data and research methods already connected. You can continue from a research question into position analysis, monitoring, and paper-trading tests. Live trading requires a broker connector with the relevant tools and permissions.
Profile holds Agent-wide preferences; memory stores facts and notes; a Playbook defines one portfolio's strategy. You still need to configure recurring tasks, keep rules current, and check results.
## When a general AI tool fits
A general AI tool can be a good choice when your work spans several domains, you already have trusted data integrations, or you need flexible coding and document creation. Compare the actual capabilities of your plan rather than assuming the product has no memory, data access, or automation.
For either path, inspect sources, quote times, assumptions, and authorization settings. Neither a data connection nor a reusable workflow guarantees correct analysis. See [connector permissions](/concepts/connectors#choose-a-permission-mode) before enabling external actions.
## Related
* [Official ChatGPT comparison](https://driven.ai/comparisons/chatgpt-vs-driven)
* [Driven vs Claude Code](/why-driven/driven-vs-claude-code)
* [What is Driven](/get-started/what-is-driven)
* [Data and coverage](/concepts/data-and-coverage)
# Driven vs Claude Code + financial data
Source: https://docs.driven.ai/why-driven/driven-vs-claude-code
Compare a ready investment workspace with configuring Claude Code, financial services, and your own research workflows.
Claude Code can use MCP integrations, read and write files, run code, and support reusable workflows. It also has [persistent project memory](https://code.claude.com/docs/en/memory) and [scheduled work](https://code.claude.com/docs/en/scheduled-tasks). It can support investment research when the required data and methods are configured.
Driven provides a dedicated investment workspace with built-in financial data, Skills, paper accounts, and scheduled tasks. Both approaches can be useful; compare setup, data rights, workflow maintenance, and account requirements.
## Compare the work you need to configure
| Area | Claude Code with financial integrations | Driven |
| - | - | - |
| Data | Choose services and obtain any required subscriptions or authorizations | Built-in coverage; add connectors for external data or brokerage tools |
| Methods | Define or install Skills, scripts, source checks, and output conventions | Pre-built investing Skills; custom and community Skills can extend them |
| Memory | Project instructions and memory files | Profile and memory within an Agent, with portfolio-specific Playbooks |
| Scheduled work | Configure an available scheduling mode and execution environment | Configure a task's prompt, time, timezone, and delivery within the Agent |
| Accounts | Depends on the integrations you add | Paper accounts are built in; live trading depends on a supported broker connector |
## Compare total requirements, not only the AI subscription
Check model usage, data licenses, external service fees, setup effort, and maintenance for either route. Claude Code does not always require per-token API billing; your access method matters. Driven's subscription has usage and feature limits, and connected services can require their own authorization or fees.
## Which path fits
Choose a configurable toolchain when you want control over data providers, scripts, and execution environments and can maintain the process. Choose Driven when you want a workspace already organized around investment research, options, portfolio context, and recurring monitoring.
For both, verify financial inputs and tool results. A stored strategy does not establish correct execution. Review [connector permissions](/concepts/connectors#choose-a-permission-mode), especially for unattended or messaging-channel workflows.
## Related
* [Driven vs general AI](/why-driven/driven-vs-chatgpt-claude-gemini)
* [What is Driven](/get-started/what-is-driven)
* [Skills](/concepts/skills)
* [Connectors](/concepts/connectors)
# Driven vs Fiscal.ai
Source: https://docs.driven.ai/why-driven/driven-vs-fiscal-ai
Compare Fiscal.ai financial statements, segments, KPIs, and APIs with Driven research, portfolio context, and recurring investment workflows.
Fiscal.ai provides a fundamentals terminal and APIs for financial statements, ratios, business segments, and operating KPIs, along with analytical Skills. Driven places data and methods within an ongoing Agent workspace connected to strategies, portfolios, and scheduled checks.
## Compare your workflow needs
| Dimension | Fiscal.ai | Driven |
| - | - | - |
| Data focus | Fundamentals, business segments, and operating KPIs | Research data across prices, statements, options, ownership, and macro |
| Access | Terminal research or APIs for your own systems | Conversation, Skills, and connectors |
| Organization | Data retrieval, comparisons, and modeling | Agent context, portfolio rules, and follow-up reviews |
## A practical workflow
After earnings, inspect statements, segments, and KPIs. Bring evidence you are authorized to use into Driven, update assumptions, and check the selected portfolio's rules. Confirm a working connector or file input before relying on external data; do not assume an automatic integration between the products.
## Requirements and fit
Evaluate Fiscal.ai for company-specific KPIs, financial modeling, or data APIs. Evaluate Driven for connecting research, positions, and recurring reviews in one workspace. Compare data depth for the actual companies and fields you need.
Features, regions, subscriptions, and data rights can change; verify your current account. Live trading needs supported broker tools and authorization. Scheduled and messaging-channel workflows can run connector tools automatically; review [permissions](/concepts/connectors#choose-a-permission-mode). Research output is informational, and investment decisions remain yours.
## Related
* [Full comparison on Driven](https://driven.ai/comparisons/fiscalai-vs-driven)
* [What is Driven](/get-started/what-is-driven)
* [Data and coverage](/concepts/data-and-coverage)
* [Official product information](https://docs.fiscal.ai/docs/introduction)
# Driven vs Google Finance
Source: https://docs.driven.ai/why-driven/driven-vs-google-finance
Compare Google Finance market and portfolio tools with Driven research workflows, Skills, portfolio Playbooks, and scheduled monitoring.
Google Finance brings quotes, news, watchlists, portfolio views, and AI research into a familiar market entry point. Driven organizes work around Agents, Skills, and portfolio rules. Google Finance also supports scheduled work, so automation alone does not distinguish the products.
## Compare your workflow needs
| Dimension | Google Finance | Driven |
| - | - | - |
| Daily use | Understand market changes and portfolio performance | Continue research, analysis, and review for personal goals |
| Methods | Use AI research and market explanations | Choose investing Skills and retain context and assumptions |
| Portfolio rules | Check available portfolio features and coverage | Define a paper portfolio's Playbook and schedule checks |
## A practical workflow
After a holding reports earnings, inspect market moves and disclosures, then ask Driven to check growth, valuation, sizing, and earlier assumptions for the selected portfolio. Define a schedule, timezone, and notification conditions for future checks.
## Requirements and fit
Evaluate Google Finance for convenient market and portfolio information. Evaluate Driven for reusable methods, stored strategy rules, and recurring reviews. Driven's specialist equity coverage focuses on the US, Hong Kong, and A-shares; this does not establish broader global coverage than Google.
Features, regions, subscriptions, and data rights can change; verify your current account. Live trading needs supported broker tools and authorization. Scheduled and messaging-channel workflows can run connector tools automatically; review [permissions](/concepts/connectors#choose-a-permission-mode). Research output is informational, and investment decisions remain yours.
## Related
* [Full comparison on Driven](https://driven.ai/comparisons/google-finance-vs-driven)
* [What is Driven](/get-started/what-is-driven)
* [Data and coverage](/concepts/data-and-coverage)
* [Official product information](https://blog.google/products-and-platforms/products/search/google-finance-updates-june-2026/)
# Driven vs Perplexity Finance
Source: https://docs.driven.ai/why-driven/driven-vs-perplexity-finance
Compare Perplexity Finance research and evidence outputs with Driven Agent context, investing Skills, portfolio rules, and scheduled reviews.
Perplexity Finance emphasizes financial research with traceable evidence and usable outputs. Driven connects research to Agent preferences, portfolio rules, scheduled checks, and position workflows. Both can support multi-source research.
## Compare your workflow needs
| Dimension | Perplexity Finance | Driven |
| - | - | - |
| Starting point | Retrieve and synthesize evidence for a question | Continue within Agent goals, positions, and research history |
| Methods | Use research tools to produce answers and deliverables | Reuse valuation, screening, and risk methods through Skills |
| Follow-up | Inspect sources and continue the investigation | Refresh data, check portfolio rules, and schedule reviews |
## A practical workflow
For a software company's margin change, collect filings, calls, and other evidence, then ask Driven whether the company fits the selected portfolio's cash-flow, valuation, and concentration rules. Store research facts in memory, portfolio rules in its Playbook, and explicitly schedule follow-up checks.
## Requirements and fit
Evaluate Perplexity Finance for broad evidence discovery and research deliverables. Evaluate Driven for recurring research and account workflows organized around your investment rules. Existing research material can also be used alongside Driven.
Features, regions, subscriptions, and data rights can change; verify your current account. Live trading needs supported broker tools and authorization. Scheduled and messaging-channel workflows can run connector tools automatically; review [permissions](/concepts/connectors#choose-a-permission-mode). Research output is informational, and investment decisions remain yours.
## Related
* [Full comparison on Driven](https://driven.ai/comparisons/perplexity-finance-vs-driven)
* [What is Driven](/get-started/what-is-driven)
* [Data and coverage](/concepts/data-and-coverage)
# DrivenBench: Investment-agent model benchmark
Source: https://docs.driven.ai/why-driven/drivenbench
Understand how DrivenBench evaluates investment-agent models and how to read capability scores, cost, latency, and regression checks.
**DrivenBench** evaluates AI models on investment-agent workflows inside Driven. It measures financial-data handling, tool use, research, and task completion to help you compare capability, cost, and waiting time.
Open the [DrivenBench leaderboard](https://driven.ai/drivenbench) for the interactive chart and model details. This page explains the metrics; model rankings belong to the version displayed on the live page.
## What it covers
| Capability area | What is examined |
| - | - |
| Data and calculations | Numerical reasoning, financial data tools, programmatic analysis, and missing-data handling |
| Research and evidence | External research, citations, evidence, and verification of delegated work |
| Files and execution | Bash execution, file creation, long documents, and large tables |
| Ongoing workflows | Scheduled tasks, trading and accounts, memory and sessions, and Skills and routing |
Separate regression checks examine failure modes such as lost citations, inaccurate coverage reporting, or data-destructive task operations. They are disclosed separately from capability scores.
## Current public methodology
The current public snapshot is **1.0**, published **August 25, 2026**, with a report date of **August 15, 2026**:
* **11 models** evaluated.
* **55 capability evaluations**, each run **three times per model**, for **165 capability runs per model**.
* **19 regression checks** reported separately, **excluded from capability scores and default ranking**.
* Capability score is the equally weighted average pass rate across capability evaluations. Equal scores share rank.
For example, `155 / 165` capability passes gives approximately `93.9%`. This describes results on this test set, not the success probability of every real task.
## Read the leaderboard
| Metric | Meaning | What to check |
| - | - | - |
| **Capability score** | Capability evaluation pass performance | Default ranking basis; inspect dimensions relevant to your task |
| **Capability passes** | Passed runs / total capability runs | The current denominator is 165 per model, excluding regression checks |
| **Cost upper bound** | Total cost upper bound for running the full capability suite three times, using the published pricing basis | Excludes cache discounts; it is not one-question pricing, a subscription price, or your current bill |
| **Latency median / p90** | Median and 90th-percentile observed evaluation runtime | Shows typical and slower runs; it is not a response-time commitment |
Models ran in different waves and provider load varied, so latency is directional. Pricing bases and model versions can also change. Read the notes attached to the leaderboard.
### Chart and model details
The chart's horizontal axis is **total benchmark cost upper bound**, on a logarithmic scale. The vertical axis is **capability score**. Further left means lower cost; higher means a higher score on this evaluation.
The solid line is the observed **Pareto frontier**: within this sample, no other model matches or improves both score and cost while strictly improving at least one. It considers those two metrics, excluding latency, your task requirements, and account access.
Select a chart point or expand a model's **Details** to inspect dimension results and disclosed regression issues. Mobile layouts also offer **Capability**, **Cost**, and **Latency** sorting; the desktop table displays these metrics together.
## Use it to inform model selection
1. Define the task: research, calculations, files, or recurring work.
2. Inspect relevant dimensions and regression issues alongside the overall score.
3. Compare cost and median/p90 runtime against your usage and waiting-time needs.
4. Try representative tasks in Driven and verify data, citations, and tool results.
Evaluation does not establish availability in your account. Access depends on your [plan](/get-started/plans-and-pricing), region, and current model versions. Benchmark costs do not directly translate into your plan usage.
Results describe specific versions, tasks, and conditions. A higher total score does not eliminate evidence, data, or execution errors. Passing regression checks does not establish safety in every real scenario. The benchmark does not measure investment returns or constitute investment advice.
## Inspect public data
* [Interactive leaderboard](https://driven.ai/drivenbench)
* [Complete JSON dataset](https://driven.ai/drivenbench/data.json) — methodology, models, dimensions, and disclosed regression events
* [Text benchmark summary](https://driven.ai/drivenbench/llms.txt)
* [Why DrivenBench?](https://driven.ai/whats-new/blog/most-expensive-ai-model-isnt-always-the-best)
When citing results, include the public version, model name and version, metric definition, and report or evaluation date. Avoid presenting an older snapshot as a live ranking.