What is MCP?
The Model Context Protocol (MCP) gives AI clients a standard way to discover tools, read resources, and run workflows over a single endpoint. FinanceDataAPI ships an MCP server so agents can look up symbols, compute technical signals, and submit backtests without custom glue code.
FinanceDataAPI mounts the server at .../v1/mcp and wraps it with the same API-key authentication, rate limiting, and monthly quota rules used by the REST API. The server currently exposes:
- Tools for symbols, signals, and backtests
- Resources for symbol browsing and signal definitions
- Prompts that guide common analysis flows
Typical flow: connect Claude Desktop, Cursor, or another MCP client to FinanceDataAPI, pass your API key as an HTTP header, and ask a natural-language question like "Compare RSI, MACD, and SMA for AAPL over the last 90 days." The client turns that into tool calls automatically.
Why use the MCP server instead of raw REST calls?
- One protocol for many clients. Claude Desktop, Cursor, Continue, and custom agents all speak the same tool/resource/prompt model.
- Structured discovery. Agents can list tools and inspect input schemas before they call them.
- Better multi-step workflows. Prompts such as
analyze_tickerandbacktest_strategyencode a recommended call order so agents do not have to infer it. - No duplicate auth model. The server accepts the same FinanceDataAPI key you already use elsewhere.
Current capabilities
| Surface | What it covers |
| --- | --- |
| Tools | list_symbols, get_symbol, list_signals, run_signals, 8 per-signal tools, and 3 backtest tools |
| Resources | symbols://list, symbols://{ticker}, signals://definitions |
| Prompts | analyze_ticker, compare_strategies, backtest_strategy |
Endpoint
https://<your-financedataapi-host>/v1/mcpFor local development, the equivalent endpoint is typically:
http://localhost:8000/v1/mcp