hindcast

Use cases & endpoints

Financial data for LangChain agents (tools + citations)

Most LangChain projects that touch markets start the same way: a custom tool per data source, a scraper behind it, and a retry wrapper bolted on when the rate limits bite. It works for a demo and becomes the maintenance burden the moment two sources disagree about what a company is called.

Hindcast is designed to be the one tool provider behind that agent. Because MCP is supported natively by the ecosystem's tool adapters, the same remote endpoint that serves a chat assistant can back a LangChain agent — with the identifier graph doing the work of making a filings result and a price series refer to the same company.

The property that matters most for agent reliability is citation. Every response carries the source, identifier and timestamp behind the value, so an agent's chain-of-thought can be checked against something concrete instead of being trusted because it sounded specific.

hindcast — session

preview

hindcast> agent asks for a fundamental figure with an as-of date

→ tool: fundamentals { ticker:"...", metric:"...", as_of:"..." }

value plus the filing and date it came from

source: preview — endpoints open to the waitlist first

Preview of the shape of an answer — endpoints open to the waitlist first.

What's covered

Join the waitlist for early access.

Questions

Do I need MCP specifically, or will plain HTTP work?

Either. MCP lets a client discover tools at connect time, which is convenient for agents; REST is there for pipelines and for frameworks where a plain HTTP tool is simpler to reason about.

How does this reduce hallucination?

It cannot stop a model from inventing text, but it removes the most common cause of confidently wrong financial answers: an ambiguous number with no provenance. When every value arrives with its source and date, wrong answers become checkable.

Join the waitlist. Vote the roadmap. First in, first served.

← Back to the data catalog