Skip to main content
This is the Knowledge plane over MCP: the tools an AI client uses to read the team’s context before acting and to write a clean Note after learning. For setup, authentication, and the OAuth scopes referenced below, see MCP & OAuth. Every read carries a flat trust field (reviewed = Knowledge, the team’s source of truth; proposed / draft = a Note, a hint), so a client can key on trust === "reviewed" directly.

Retrieval and reading

Start with driftless_context_retrieve when you have a task but no slug. Bodies come back brief by default (the durable why, not the full content); fetch one full body with driftless_context_get. Permission: context:read. Availability: Available.

Writing and editing

Permissions: topics:create / topics:write; work:write for merge and share. Availability: Available. Write responses carry anchor_validation (ok / overbroad / zero), which never blocks the write.

Graph and health

Permission: context:read. Availability: Available.

Registries: tags and areas

Permissions: list is member-level; create / update / delete need work:write. Availability: Available.

Governance

A topic becomes Knowledge only once an owner or admin merges it. An MCP client always proposes; it can merge only with owner/admin authority, and only when the authorizing person explicitly asks. Availability: Available. Automatic merge or auto-approval is Not available: a person always decides.

Permissions

Merging into Knowledge additionally requires owner/admin authority, not just a scope.