API
Everything the app does goes through one HTTP API, which connected agents use too. The complete reference for agents is the skill; this page is the outline.
Requests
JSON in and out under /api/v1, with a key as Authorization: Bearer <key>. A refusal is {"error": "code", "message": "a sentence for a person"}; pass the message on as it is.
| Request | What it does |
|---|---|
GET /api/v1/you | your documents, requests, keys and limits |
POST /api/v1/documents | make a document, with its first version |
GET /api/v1/s/{doc} | open a document: its drafts, version, pages and comments |
GET /api/v1/s/{doc}/file?version=&name= | one page or file, as stored |
GET /api/v1/s/{doc}/versions?draft= | a draft's versions, newest first; draft=all for every draft's |
POST /api/v1/s/{doc}/save | save the next version of a draft |
POST /api/v1/s/{doc}/drafts | begin a draft |
POST /api/v1/s/{doc}/drafts/{draft}/merge | merge a draft back where it began |
GET /api/v1/s/{doc}/compare?from=&to= | what each page did between two versions |
GET, POST /api/v1/s/{doc}/comments | read the comments, or write one |
POST /api/v1/s/{doc}/access | ask somebody already on the document for feedback |
GET /api/v1/s/{doc}/events | live updates, as server-sent events |
GET /api/v1/you/search?q= | the documents you can read that say these words, best first |
MCP
POST /mcp speaks the Model Context Protocol with the same key. Its tools are find_documents, open_document, read_file, list_versions, list_comments, comment, resolve_comment, write_version, update_document, image_upload_link, compare_versions, merge_draft, ask_approval and ask_for_feedback; each is one of the requests above.
For agents
/llms.txt lists these pages as Markdown, and /llms-full.txt is all of them in one file.