Skip to main content

Glean in Gemini chat

This admin guide covers connecting Glean to Google Gemini Enterprise using the Glean Model Context Protocol (MCP) server, so end users can ask work questions in Gemini and get answers grounded in company knowledge from Glean.

Glean in Gemini connects your organization’s permission-aware enterprise knowledge to Gemini Enterprise through a custom MCP server data store. Once connected, Gemini can search, read, and synthesize across the systems Glean already indexes.

Gemini chat answering a work question with grounded citations from Glean

With the recommended default tool set, this integration is read-only — it can search and synthesize company knowledge but does not create, update, or delete records in connected systems. If you enable write-capable tools on the selected MCP server path, Gemini can surface those actions as well.

What users get

  • Ask Gemini questions that need company context and get answers grounded in Glean.
  • Receive answers with citations to the underlying internal sources.
  • Use the Glean tools enabled on the selected MCP server path as Gemini actions.

How it works

Glean exposes a Model Context Protocol (MCP) server. You register that server in Gemini Enterprise as a custom MCP server data connector. When a user chats with Gemini, Gemini calls Glean's MCP server, authenticates the user via OAuth, and uses the returned context to answer.

Gemini chat ──▶ Custom MCP Server connector ──▶ Glean MCP server
(OAuth per user) (permission-aware)

Components

  • Glean MCP server: Exposes Glean's search and knowledge tools over a standard protocol that Gemini can call. The default path (/mcp/default) loads 12 actions: search, read_document, chat, code_search, employee_search, user_activity, memory, gmail_search, meeting_lookup, knowledge_graph_query, knowledge_graph_schema, and memory_schema.
  • Glean OAuth client: Lets Gemini authenticate each user to Glean, so results are always returned with that user's permissions enforced.
  • Gemini data store and app: Registers the Glean server inside Gemini Enterprise and surfaces it to end users through a branded web app.

Core capabilities

CapabilityWhat it doesVisibility
Grounded answersGemini answers work questions using company knowledge retrieved from Glean.Each answer is generated for the asking user and reflects their Glean permissions.
CitationsAnswers include links to the underlying internal sources in Glean-connected systems.Citations are personalized — each user only sees sources they can already access.
Tool invocationGlean tools enabled on the selected MCP server path are exposed as Gemini actions (for example, search, read_document, chat, code_search, employee_search, user_activity, and memory).Tools run on behalf of the signed-in user and respect Glean permissions.
Permission-aware authenticationEach user authorizes Glean via OAuth from inside Gemini.Each user authorizes individually; every response respects that user's Glean permissions.

Get started

Who is involved

This setup involves two admin personas and one end-user persona.

PersonaResponsibilities
Glean adminProvide the Glean MCP server URL, confirm the required MCP tools are enabled, and create the Glean OAuth client.
Google Cloud or Gemini Enterprise adminCreate the custom MCP server data store, enable actions, connect the data store to a Gemini app, and publish the app.
End userAuthorize Glean the first time they use the app, then ask questions in Gemini chat.

Requirements

  • Glean admin access — to locate the MCP server URL and create an OAuth client.
  • Google Cloud or Gemini Enterprise admin access — to create the data store, enable tools, and publish the app.
  • A Google Cloud project — Gemini Enterprise resources are created inside a project you own.
  • Your Glean backend URL (<tenant_id>-be.glean.com) — used to build the server, authorization, and token URLs. Find it in About Glean under Server instance (QE).

For the full list, including details you will collect during setup, see Prerequisites.