MCP servers
Financial data MCP server — one endpoint for market data
Most agents that touch financial data end up wired to a handful of narrow integrations — one for filings, another for prices, a third scraped from a statistics agency — each with its own identifiers, auth and failure modes. The glue code becomes the product, and it rots.
Hindcast is a single remote MCP server across the catalog: SEC and EDGAR, Brazilian market and fund data, central bank communication, point-in-time macro, institutional holdings, transcripts, entity resolution, positioning, prediction markets and commodities. One connection, one identifier space, one citation convention. An agent that can resolve a company can immediately ask about its filings, its price history, who disclosed holding it and what management said on the last call.
Two properties are deliberate. Everything is point-in-time, so an as-of question returns what was knowable then. And every answer carries its source, so a claim can be verified rather than trusted. Endpoints open to the waitlist first, and the catalog is built in the order the waitlist votes.
hindcast — session
previewhindcast> connect an MCP client and list the available tools
→ tool: mcp_list_tools { }
✓ one tool family per catalog dataset, all sharing identifiers
source: preview — endpoints open to the waitlist first
What's covered
- One remote MCP endpoint spanning all eleven catalog datasets
- Shared entity identifiers across filings, prices, holdings and text
- As-of arguments on historical tools, so vintages are the default
- Citations — source, identifier and timestamp — on every response
- REST with OpenAPI alongside MCP for production pipelines
- No SDK required; any MCP-capable client connects with a URL
Join the waitlist for early access.
Questions
Why MCP rather than a REST-only API?
Because an agent should not need a bespoke integration per data source. MCP lets a client discover the available tools at connect time, which is how one endpoint can grow new datasets without every caller shipping an update. REST remains available for pipelines that want it.
What does one endpoint actually buy me?
Shared identity, mostly. The expensive part of multi-source research is not fetching data, it is agreeing that two records describe the same company — and that agreement is what a single graph provides for free.
Is any of this live today?
Endpoints open to the waitlist first, and no dates are published. What is real right now is the catalog and the demand board: the votes determine what gets built next.
Related
Join the waitlist. Vote the roadmap. First in, first served.