When you start doing research on startups, you always start tabbing over to Crunchbase to get funding information, CB Insights for market information, PitchBook for financials, and then you use the open web. Finally, you have to reconcile all these sources on your own.
So I decided to build a thing that can query all 4 sources at once, rank the responses, and give you a ranked list while still including the confidence score for each source. The public code is on GitHub at 4ugusta/ai_intelligence_mcp. The code is stubbed out with connectors so you can run the code without needing a data license. The rest of this is not a plan to build this, but the code itself.
What "MCP" means here
MCP, for Multi-source Company Profiler, is a FastAPI server with only 1 endpoint (POST /query). This is not a Model Context Protocol server and it has no models in the request path. The last section shows what it would take to build one.