Agent-first by design

Finance for AI

Give AI agents structured access to financial knowledge — stable entities, typed relationships, honest provenance.

Why agents need a foundation layer

AI agents are becoming first-class consumers of financial information — and most financial websites are unreadable to them: data trapped in charts, paywalls and PDFs. Finance Foundation is built agent-first: every page has an API twin, entities have stable ids, and provenance is explicit so an agent knows exactly how fresh a number is.

AI financial analyst

Ground analysis in typed entities and real relationships instead of scraped text.

Portfolio research

Resolve holdings to entities, then traverse: fund → manager → listings → jurisdictions.

Company intelligence

"Which company is this?" answered with identifiers, peers and venue context.

Market intelligence

Venue and jurisdiction data with MIC codes — the where of every instrument.

Due diligence

Entity + graph + research briefs as structured context for deeper investigation.

Automated research

GET /api/v1/all — the full dataset in one call, sized for a context window.

How to point an agent here

Machine-readable index

GET https://finance.foundation/llms.txt # describes every dataset and endpoint, # in plain text an agent can read first

One-call context

GET https://finance.foundation/api/v1/all # entire dataset, one JSON response

Example tool definition

{ "name": "finance_foundation_lookup", "description": "Look up structured data on companies, markets, assets, funds, banks and countries. Reference data, not real-time.", "input": { "entity_type": "companies|markets|assets|…", "id_or_query": "string" } // call GET /api/v1/{entity_type}/{id} // or GET /api/v1/search?q={query} }

Honesty contract for agents

Every response carries data_version. Figures are approximate reference values (seed-0.1 · 2026-08-28), not real-time prices — correct uses are entity resolution, relationships, classification and context, not live quoting. An agent that needs live prices should combine this layer with a market-data feed.

Roadmap: an MCP server exposing these endpoints as native agent tools, and usage-based agent access. Early interest: admin@digitaldomains.market.