Internal knowledge service
Company knowledge, retrievable.
Corpus ingests our documents, indexes them for hybrid search, extracts structured facts about clients, projects and people, and serves all of it to other systems over MCP.
Ingest
Documents arrive from connected sources, are parsed into sections, and classified by kind.
Index
Hybrid retrieval over chunks, sections, and whole documents: vector, keyword, and rerank.
Extract
An extraction pass proposes clients, projects, and people. Nothing is confirmed without an operator.
Serve
Read-only tools over MCP, so assistants and internal tools answer from the same source.
Every answer carries its provenance
Search results and extracted facts point back to the document they came from: title, kind, source URL, and page range. Nothing is asserted without a trace.
Operators review proposed entities, merge duplicates, govern the document-kind vocabulary, and publish organization-wide guides from the admin console.
MCPRead-only tools for assistants/internal/searchHybrid search over the corpus/adminOperator console, SSO only/assets/:idPublic asset pass-through
Access is limited. Corpus holds internal company material. Sign-in is restricted to our Google Workspace domain, and API clients use individually issued tokens.
Sign in