Open standard · v0.1.0 · live
Put your company info so AI agents can read it easily.
One small JSON file at a known location on your own website. AI agents and the open directory discover it without scraping, login walls, or paying a data broker. You stay the source of truth - for everything, from the public-facing pitch all the way through to the figures behind NDA.
Why this exists
Three things are happening at once, and the existing standards stack doesn't meet any of them.
Agents have become the reader
LLM-driven agents now do the first pass on every diligence query, RFP, sourcing list, and supplier check. They need machine-readable answers, not HTML.
The web has no company layer
Schema.org describes content. OpenGraph describes social previews. There's no open standard for "who is this company, and what do they want a serious reader to know."
Brokers own the data we publish about ourselves
PitchBook, Crunchbase, and others assemble company profiles by scraping and licensing - then sell them back to investors at five-figure annual seats. The company being described doesn't control any of it.
What it is
A single JSON file at /.well-known/agentic-profile.json
on your own domain. Two tiers, one shape:
-
Public tier - identity, jurisdiction, registry IDs,
banded metrics, evidence links, contact channel. Crawled by the
open directory. Banded numbers (
"11-50 staff","1m-5m raised") keep it clear of UK FCA financial-promotion rules so it can stay open. - Protected tier - precise figures, fundraise detail, named logos, references, calendar. Served from your own private MCP, behind your own auth, to principals you've explicitly authorised. The directory never sees it.
Same shape, two audiences. The publisher controls who sees what.
Why companies should take control
Today, the canonical "company profile" most agents and investors read is whichever broker they happen to subscribe to. That broker decides what's correct, what's missing, what's stale, and who can see it. The company has no edit access, no canonical URL, and no way to mark a claim as evidenced versus assumed.
The right place to publish data about your company is your own website. That's true of your prices, your team, your case studies, and your blog - and it should be just as true of your jurisdiction, your registry IDs, your stage, your headcount band, your fundraise window, and the contact channel you actually want agents to use.
| Question | Today | With agentic-first |
|---|---|---|
| Who's the source of truth on your company? | A data broker that bought your details from somewhere | You are, at /.well-known/agentic-profile.json |
| How does an investor agent reach the right person? | LinkedIn DM, scraped email, contact form lottery | The channel you nominated in contact.preferred_channel |
| Where do diligence-grade figures live? | Spreadsheet in a data room, gated by manual NDA | Your own protected MCP, OAuth-scoped per section, audited |
| Who pays to discover you? | The investor (broker subscription) | Nobody - discovery is open and free at the point of use |
| How is "verified" defined? | Whatever the broker decides, opaque | Statutory IDs (Companies House, GLEIF, EDGAR) the directory checks publicly |
Who it's for
Founders & IR leads
Take canonical control of how your company appears to agents and investors. Five fields and a JSON file gets you indexed; a private MCP later gets you a deal-grade surface, on your terms.
NEDs, advisors, fractionals
Publish a personal profile in the same shape - current roles, banded past-roles count, credentials, evidence - on your own site. Discoverable without LinkedIn.
Investors & operators
Filter the open directory by jurisdiction, stage, funding band, headcount band, score. Every result links back to the publisher's source URL - no scraping, no broker tax.
Agent & tool builders
One Streamable HTTP MCP, no auth required for read. Drop it into Claude Desktop, Cursor, or your own client. Start here →
All the technical detail lives on GitHub
This is the consumer page. The full written spec, the four canonical JSON Schemas, copy-paste publishing recipes for every common host, the security policy, drop-in skills for Claude and Codex, the Python validator, and the relationship to LEI / XBRL / Verifiable Credentials all live in the open repo:
AI agents pointed at this domain should read /llms.txt for a structured index.