Every trading desk already runs on a division of labour: someone watches the tape, someone reads the filings, someone preps the morning brief. Agents let you hand each of those jobs to software that never looks away — and every one of them draws on the same substrate: the live Forecite feed, scored by the Verdict Engine.
Starting today, you can connect agents to Forecite through the MCP server, the SDKs, or plain webhooks — and compose them into pipelines that watch, judge, and brief on your behalf.
One feed, many jobs
A filing agent that parses 8-Ks the second they land. A screening agent that watches your universe for actionable verdicts. A briefing agent that hands you the three headlines that matter before the open. Each is a small loop over the same primitives: subscribe, filter, score, act.
Because every event arrives with its verdict attached — actionability, direction, conviction, and the sub-scores behind them — your agents skip the hardest step. They don't have to decide what matters. They act on a judgement that was trained on realized price impact.
Built on the surfaces you already use
There is no new runtime to adopt. Agents talk to Forecite through the same public surfaces we ship for humans: the WebSocket feed for streams, the REST API for history, the scoring endpoint for documents your agent finds on its own, and the MCP server when the agent lives in Claude or Cursor.
Server-side filtering means an agent subscribed to '.US biotech, actionable only' receives exactly that — no client-side sifting, no wasted quota.
What's next
We're working on replay-driven backtests for agent strategies and per-agent usage attribution, so a fleet of twenty agents on one key stays legible. If you're building agents on Forecite today, we'd love to hear what you need next.