hadi-labs — a working preview.

three jobs an AI agent can already do inside an investment firm

Three agent surfaces paired to how an investment firm actually runs its workflow: thesis kickoff on a new name, inbound deal triage, and holdings monitoring. Backed by Claude Sonnet, the Anthropic Managed Agents runtime, and FMP for live financial data. A labs-style proof of what's buildable in-fund — not a finished product.

Start here

Thesis kickoff

I want to see how the agent picks up a new stock idea and turns it into a fund-shaped memo — with prior-evaluation memory recall, portfolio overlap check, IC criteria scoring, and citation discipline.

Works for: US-listed public equities. Three curated tickers (GRAB / AXTI / IHH) have seeded memory + hand-curated narrative; any other ticker uses the attached FMP MCP server for live data.

Try kickoff →

Deal triage

I want to see how the agent reads an inbound deal teaser or pitch memo, extracts the structured fields, checks portfolio overlap, scores against the fund's IC criteria, and recommends pursue / monitor / pass.

Works for: both public stock pitch memos AND private deal teasers (Series A through later-stage growth).

Try triage →

Anomaly

I want to see how the agent explains a material move on a holding — looking up the original thesis, scanning news, comparing peer moves, and classifying the move as thesis-intact, questionable, or thesis-breaking.

Works for: any held position (stocks or privates) when something material happens — price drawdown, mark-down, sector-wide event.

Try anomaly →

where each surface is going →