Guru's value proposition is that answers are verified by someone accountable. An agent that could publish directly would undermine exactly that. Create Draft and Update Card as suggestions keep the human verification step intact while still letting the agent do the work of writing. The logging matters too: every MCP interaction lands in Guru's own activity record, so an unexpected answer can be traced.
Guru's endpoint over their knowledge platform. Five capabilities, named in their documentation by friendly name rather than wire-level tool name.
- List Knowledge Agents — the Guru agents available in your workspace
- Ask — a question answered with Guru's verification behind it
- Search — cards and connected documents matching a keyword
- Create Draft — a draft card straight from the conversation
- Update Card — suggest changes to an existing card
OAuth 2.0 with PKCE and dynamic client registration, or an API token over basic auth. Responses are restricted to the authenticated user's RBAC permissions, and every MCP interaction is logged in Guru's AI activity record.
Paste a URL, then authorize — add the endpoint to your client, then approve the OAuth consent screen