Use cases
The common thread: something is about to be decided about a company nobody has actually looked at. At 12 credits a dossier, looking is cheaper than guessing — and far cheaper than acting on a guess someone else made.
Autonomous sales agents
An agent working a list of domains needs to know which ones are worth writing to. Industry, an idea of size, a stack that says whether you integrate with them — and, crucially, an honest blank where the company simply hasn't published enough to judge.
How: build_company_dossier over MCP, one call per domain, in the qualification step.
sections: identity, size, stack
Outbound and lifecycle
The difference between a message that lands and one that gets deleted is a specific, checkable detail. Every field in a dossier arrives with the URL it was read from, so the sentence you write can point at the page you read it on.
How: Store the source URL beside the field; use it as the citation in your own CRM notes.
sections: identity, funding
Partnerships and integrations
Whether a prospect is on Shopify, ships a Next.js app, runs HubSpot or has a Greenhouse board changes the pitch entirely. The stack comes from response headers and the assets their homepage actually loads, with the header line or asset URL attached to each detection.
How: Read tech_stack; each item carries evidence and the kind of signal it came from.
sections: stack
Recruiting and market mapping
Live job links on a careers page are the most honest growth signal a company publishes. Counted deterministically from ATS links and job paths — and returned as null, never zero, when we could not tell.
How: careers_open_roles plus headcount_hints; treat null as unknown, not as none.
sections: size
Diligence and vendor review
Before a call, you want the description they give themselves, whether anyone has written about their funding, and how to reach a human. What you do not want is a confident industry label invented from the domain name.
How: Read dropped_claims: it tells you how much the model tried to add that the pages did not support.
full dossier
CRM and data teams
The problem with enrichment data is that six months on, nobody can say where a field came from or whether it was ever true. Here every value carries its source URL, so a stale field can be re-checked against the page it came from.
How: POST /api/dossiers, persist the source alongside every value you keep.
full dossier
Company Intel reads what a company has published about itself and what has been published about it. It cannot tell you revenue, headcount, churn or who the decision maker is unless the company wrote it down somewhere public. On all of those it returns an empty field — and if the domain holds nothing usable at all, it refunds you for the privilege of hearing so.