Most memory features are write-and-hope: you cannot see what shape the store expects or what its limits are. Publishing the categories, payload schemas, filterable fields and storage limits as a discoverable tool means an agent can write memories that will actually be retrievable later. On the search side, permission-aware results through the knowledge graph is the whole reason enterprises deploy Glean — an agent gets exactly the documents the person could have found themselves.
Glean's endpoint, addressed per customer at your own backend domain. Administrators can create additional servers alongside the default one, each scoped differently.
- search — across the whole knowledge base, with results filtered by what you are allowed to see
- chat — conversations with Glean's assistant
- memory and memory_schema — read, add, update and delete personal memories, and discover the categories, payload schemas and storage limits
- read_document, code_search, employee_search, user_activity
- gmail_search, outlook_search, meeting_lookup — mail, calendar and meeting recordings
OAuth 2.1 against Glean's own authorisation server. Every result enforces user-level access control through Glean's knowledge graph. Enterprise pricing, quote-based, with MCP included as a platform feature.
Paste a URL, then authorize — add the endpoint to your client, then approve the OAuth consent screen