RAG corpora
Company facts as a structured corpus for vertical AI: one record shape per company, every value dated and sourced, retrievable by company, dimension, or whole category.
Structured facts, not scraped prose. Chunking is your call, not ours.
The job
Most company corpora are a pile of pages.
The usual build is a crawler, an HTML-to-text step, and an embedding job, which produces chunks that answer well until someone asks when a fact became true. The date lives in the page you threw away, so the corpus can only ever say what it says now.
Here the unit is a fact rather than a paragraph: typed, dated, attached to the capture it came from, and grouped by dimension so a retrieval set can be scoped to what your product actually reasons about.
What that costs today
- Chunks that cannot say when they became true
- Re-crawling the same sites every team builds on
- Prose embeddings where the answer needed a field
- No trail from a retrieved chunk back to a source
The API
4 calls cover the job.
Current state is open. History values, claim checks, and evidence trails sit behind a free key. These are the real routes.
/v1 /dimensionsWhat kinds of fact exist, and how each is sourced.
/v1 /categories /{slug}A whole market's membership, as the retrieval boundary.
/v1 /companies /{company} /factsFacts per dimension, each with provenance and as_of date.
/v1 /companies /{company}The full current record for one company in a single call.
For agents · MCP
Or let your agent run it.
Connect the Bixel MCP server and any agent, Claude, an internal copilot, a procurement bot, works this job against live structured data, with every value citing the dated capture it came from. The recommendation becomes auditable, not a guess.
You ask your agent
“Pull the pricing and security facts for every company in the vector-databases category, with the date on each.”
A retrieval set that knows when it was true.
A bounded category is the natural retrieval set; this is one, live: see the bounded set →
More in AI & Agents
The answer layer for agents that need company facts with receipts.
Agent grounding over MCP
Connect the MCP server and your agent answers company questions with dated, sourced facts.
Explore
Claim verification
Given a generated sentence about a company, return supporting or contradicting facts with provenance.
Explore
Factuality benchmarks
Dated ground truth for evaluating models on what was true as of a given date.
Explore
Entity resolution
Domain to company to parent, with subsidiaries and renames handled.
Explore
Procurement agents
Agents that shortlist, compare, and monitor vendors use the record as their memory.
Explore