Skip to main content
By 2026, the Claude ecosystem has real financial-data capability. Between official connectors, community MCP servers, and Anthropic’s Financial Services marketplace, a technical user can assemble something that looks like an investing Agent out of Claude Code plus data sources. This page is for the reader evaluating that path against Driven.

​
What the Claude ecosystem offers

The pieces are genuinely capable:
  • Anthropic’s Financial Services marketplace — plugins for DCF modeling, equity research, investment banking, private equity, and wealth management
  • Official MCP connectors — FactSet, S&P Capital IQ, MSCI, PitchBook, Morningstar, LSEG, Daloopa, Moody’s
  • Community MCP servers — Financial Modeling Prep, Alpha Vantage, SEC XBRL parsing, and more
  • Claude Code itself — tool calling, MCP integration, file system access, code execution
On the surface, Claude Code plus these data sources plus the official finance plugins looks like it assembles into an investing Agent. The difference is not in any single module. It is in how they are put together.

​
Agent = Model + Harness

The framing again: an Agent is a model plus a harness. At the model layer, everyone is using something comparable, Claude, GPT. The real difference is the harness: all the runtime infrastructure around the model. Here is the comparison, layer by layer.

​
1. Out-of-the-box vs assemble-it-yourself

This is the most surface-level difference, and the most practical. The Claude Code path: install Claude Code, find the MCP servers you need, configure each one’s API key, write configuration to define behavior, design your own prompt workflows, handle data-format conversions, and build your own persistence. This is, in effect, doing your own harness engineering. The Driven path: sign up, start asking. Skills, data sources, Playbook, scheduled tasks, and paper trading all work out of the box. Claude Code’s target user is a developer. Driven’s target user is an investor. A typical investor will not, and should not have to, configure MCP servers and write Agent configuration files.

​
2. Fragmented data vs unified orchestration

Claude Code connects financial data by mounting MCP servers on demand, one for quotes, another for statements, another for holdings. Each source has its own format, field definitions, and update frequency, and the model has to understand and reconcile them itself. Driven’s 245 data endpoints are orchestrated together. When a Skill runs, it knows which source to pull which field from, how to cross-check, and how to handle missing data. This is not simply “more APIs connected.” It is a different quality of data harness. A concrete example: ask Claude Code plus a fundamentals connector to run a DCF and it can pull the data and run the model. But it will not automatically cross-check management guidance, compare against peer valuations, or flag data confidence. Driven’s Valuation Matrix does these things because that logic is built into the Skill’s workflow.

​
3. Stateless vs stateful

This is the most fundamental architectural difference. Each Claude Code run is essentially stateless. It can read and write files to “remember” things, but that memory is passive and unstructured. It will not proactively load your investment strategy into a new session, and it will not automatically connect your position changes to market events. Driven’s Agents are stateful:
  • The Playbook persists across sessions, and every analysis runs within that strategy framework
  • The portfolio continuously tracks positions, P&L, and risk exposure
  • Scheduled tasks run independently of any user session, 24/7
  • Multiple Agents each maintain their own independent strategy context
A scenario makes the difference concrete. Monday morning, you open Driven: the Agent has already run a pre-market scan against your watchlist, flagged the earnings events relevant to your holdings, and summarized the unusual volume from Friday’s close. You did not ask for any of it. The information is simply there. Claude Code cannot do this, because it has no concept of “always running” and no structured strategy and portfolio context to drive proactive behavior.

​
4. General finance plugins vs investor-first Skills

Anthropic’s Financial Services plugins are strong, but they are designed for finance professionals, investment bankers, PE, asset managers, compliance, not individual investors. A DCF plugin outputs a professional three-statement model; an equity research plugin outputs institutional-format reports. Valuable for professionals, but for an individual investor the barrier is high and the information density is impractical. Driven’s Skills are designed around an individual investor’s daily workflow. Market Pulse gives a quick market read; Stock Screener does multi-factor selection; Smart Money tracks institutional activity; Portfolio Monitor checks holding health. The output format, granularity, and action orientation are all aimed at making an investment decision, not writing a research report.

​
5. Cost and maintenance

The Claude Code path’s cost structure: Claude API usage (per token), plus each data source’s subscription (FMP, Alpha Vantage, FactSet, and so on, billed separately), plus ongoing maintenance, MCP servers can disconnect, API formats change, sources get upgraded. The total likely exceeds Driven’s subscription, and the maintenance is continuous. The MCP ecosystem is still early; service stability and data quality vary. Driven is one subscription with all data sources, Skills, scheduled tasks, and paper trading included.

​
Who should build it themselves

The Claude Code path genuinely suits developers and quantitative teams with the technical ability and the willingness to invest the time, people who want maximum control and customization and are happy to be their own harness engineers. If that is you, it is a legitimate route. Driven is for investors who want the assembled, optimized, investing-specific Agent without building and maintaining it. The model layer is comparably capable; the harness, data orchestration, Skill workflows, Playbook memory, scheduled tasks, portfolio management, paper trading, is pre-built and tuned for an investor’s daily workflow.