RFP work fails when someone pastes an approved-sounding answer that was never approved. Returning a trust score and the source alongside each result makes that checkable, and the scopes make it structurally impossible for the agent to write a new answer into the library. list_content_usage is the underrated one: knowing an answer has been reused across twelve projects tells you more about whether to trust it than any score.
AutoRFP's endpoint, published as a read-only release — the write tools are stated as future work. Three regional endpoints exist: US, EU and a default.
- search_content — semantic search across your approved content library, filtered by tag
- list_content and list_content_usage — enumerate the library, and reverse-look-up which projects reused a piece
- list_projects and get_project — RFPs, DDQs and questionnaires with requirement status counts
- list_requirements — the questions and answers inside a project, filtered by status, tag, editor or whether an answer exists
- list_tags — the organisation's tag vocabulary, so every other call can be scoped correctly
Per-user OAuth with three read scopes: tags:read, projects:read and content:read. Responses carry a trust score and source citation metadata.
Paste a URL, then authorize — add the endpoint to your client, then approve the OAuth consent screen