MCP Stock Analysis Server
Self-hosted research agent built on the Model Context Protocol.
An automation server that streams fundamentals, technicals, and macro data into AI-first trading workflows.
Stack: Python · FastMCP · PostgreSQL · Model Context Protocol
Why it exists
Capital Companion's infrastructure needed a hardened back-end for data acquisition, tool orchestration, and evaluation. The same engine now ships as an open MCP server so builders can plug any LLM into a predictable trading workflow.
What you get
- MCP-compliant tools for fundamentals, alternative data, technical indicators, and macro feeds.
- Streaming output designed for copilots and voice agents, complete with benchmark comparisons and guardrail scoring.
- Deployment blueprints for AWS Copilot, Docker Compose, and GitHub Actions so you can launch in under an hour.
Built for
Engineers who want to stand up an investment research agent, quant teams experimenting with MCP-native execution, and founders who need a trustworthy AI back-end before exposing customers to automation.
Roadmap
Upcoming improvements include first-class support for custom data vendors, agent evaluation harnesses, and benchmarking suites that plug into LangSmith, Humanloop, or custom analytics pipelines.
Focus areas
- Implements the MCP spec with tools for fundamentals, news sentiment, and technical indicators.
- Ships with guardrails, task queues, and evaluation harnesses for safe autonomous execution.
- Deploys on AWS Copilot or Docker with infrastructure as code and scheduled refresh jobs.