Skip to main content
This guide provides practical, tested examples of using Glean’s MCP server in tools like Cursor, Claude Desktop, VS Code, and other MCP hosts. These examples come from real usage patterns and have been validated in production environments.

What You’ll Find Here

Role-specific examples organized by job function, each including:
  • Common use cases: What teams actually accomplish with MCP
  • Real prompts: Copy-paste examples you can try immediately
  • Prerequisites: Required connectors and setup
  • Best practices: Tips for getting better results

Getting Started

Before diving into role-specific examples, ensure you have:
  1. MCP server configured: Use the MCP Configurator to install the MCP server in your MCP host
  2. Relevant connectors: Connect the data sources your team uses
  3. Appropriate permissions: Verify you can access the data sources you need
The effectiveness of these prompts depends on having the right data sources connected. Each role section lists recommended connectors.

Examples by Role

How MCP Tools Work

When you use Glean’s MCP server, your AI assistant can:
  • Search Glean: Find relevant documents, conversations, and code across your company
  • Read documents: Retrieve full content from specific URLs or IDs
  • Query structured data: Access databases, spreadsheets, and data warehouses
  • Find people: Look up employees and their areas of expertise
  • Access email: Search Gmail or Outlook for relevant messages
  • Chat with Glean: Ask Glean Assistant questions about your data
All MCP queries respect your existing permissions. You’ll only see data you already have access to in Glean.

Prompting Tips

To get the most from Glean’s MCP server:

Be Explicit About Tools

Tell your AI assistant to use Glean:
  • ✅ “Search Glean for recent PRs related to authentication”
  • ✅ “Use Glean to find the design doc for feature X”
  • ❌ “Tell me about authentication” (may not use MCP)

Specify Data Sources

Mention specific connectors when relevant:
  • ✅ “Search Glean for Zendesk tickets about login errors”
  • ✅ “Find Slack conversations in #engineering about this bug”
  • ❌ “Find tickets about login errors” (less specific)

Provide Context

Include relevant identifiers, dates, or keywords:
  • ✅ “Given Jira ticket EN-12345, find similar past issues”
  • ✅ “Search for discussions about the Q3 roadmap in the last 30 days”
  • ❌ “Find similar issues” (too vague)

Iterate and Refine

Start broad, then narrow based on results:
  1. “Search Glean for information about our deployment process”
  2. Review what comes back
  3. “Now focus on the Kubernetes deployment docs from the last quarter”
I