@gleanwork/local-mcp-server
v0.9.1
Published
MCP server for Glean API integration
Downloads
3,546
Readme
@gleanwork/local-mcp-server
Note: We recommend using the remote MCP server integrated directly in Glean instead of this local MCP server. The remote server provides a more seamless experience with automatic updates, better performance, and simplified configuration. This local MCP server is primarily intended for experimental and testing purposes.
The Glean MCP Server is a Model Context Protocol (MCP) server that provides seamless integration with Glean's enterprise knowledge.
Features
- Company Search: Access Glean's powerful content search capabilities
- People Profile Search: Access Glean's people directory
- Chat: Interact with Glean's AI assistant
- Read Documents: Retrieve documents from Glean by ID or URL
- MCP Compliant: Implements the Model Context Protocol specification
Tools
company_search
Search Glean's content index using the Glean Search API. This tool allows you to query Glean's content index with various filtering and configuration options.
chat
Interact with Glean's AI assistant using the Glean Chat API. This tool allows you to have conversational interactions with Glean's AI, including support for message history, citations, and various configuration options.
people_profile_search
Search Glean's People directory to find employee information.
read_documents
Read documents from Glean by providing document IDs or URLs. This tool allows you to retrieve the full content of specific documents for detailed analysis or reference.
MCP Client Configuration
To configure this MCP server in your MCP client (such as Claude Desktop, Windsurf, Cursor, etc.), run @gleanwork/configure-mcp-server passing in your client, token and instance.
# Configure for Cursor
npx @gleanwork/configure-mcp-server --client cursor --token your_api_token --instance instance_name
# Configure for Claude Desktop
npx @gleanwork/configure-mcp-server --client claude --token your_api_token --instance instance_nameFor more details see: @gleanwork/configure-mcp-server.
Manual MCP Configuration
To manually configure an MCP client (such as Claude Desktop, Windsurf, Cursor, etc.), add the following configuration to your MCP client settings:
{
"mcpServers": {
"glean": {
"command": "npx",
"args": ["-y", "@gleanwork/local-mcp-server"],
"env": {
"GLEAN_INSTANCE": "<glean instance name>",
"GLEAN_API_TOKEN": "<glean api token>"
}
}
}
}Replace the environment variable values with your actual Glean credentials.
Docker Deployment
As an alternative to npx, you can run the Glean MCP server in a Docker container. This provides isolation and consistent runtime environments.
Multi-architecture Docker images are published to GitHub Container Registry and work on both Intel/AMD (amd64) and Apple Silicon (arm64) systems.
Pull the Image
docker pull ghcr.io/gleanwork/local-mcp-server:latestMCP Client Configuration
Configure your MCP client to use the Docker image. Most MCP clients support passing environment variables via the env block:
{
"mcpServers": {
"glean": {
"command": "docker",
"args": [
"run",
"-i",
"--rm",
"ghcr.io/gleanwork/local-mcp-server:latest"
],
"env": {
"GLEAN_INSTANCE": "your-instance",
"GLEAN_API_TOKEN": "your-token"
}
}
}
}If your MCP client doesn't pass the env block to Docker, use -e flags in the args instead:
{
"mcpServers": {
"glean": {
"command": "docker",
"args": [
"run",
"-i",
"--rm",
"-e",
"GLEAN_INSTANCE=your-instance",
"-e",
"GLEAN_API_TOKEN=your-token",
"ghcr.io/gleanwork/local-mcp-server:latest"
]
}
}
}Important: The -i flag is required for stdio transport communication.
Environment Variables
GLEAN_INSTANCE(required): Your Glean instance nameGLEAN_API_TOKEN(required): Your Glean API token
Troubleshooting
Container exits immediately:
- Verify environment variables are set correctly
- Check Docker logs:
docker logs <container-id> - Ensure the
-iflag is present in the args
Permission or authentication errors:
- Verify your
GLEAN_API_TOKENis valid - Check your
GLEAN_INSTANCEmatches your Glean deployment
MCP client can't connect:
- Verify Docker is installed and running
- Check that
dockercommand is in your PATH - Review MCP client logs for error messages
Environment variables not being passed:
- Try using
-eflags in args instead of theenvblock (see alternative configuration above)
Docker Compose
For standalone deployment scenarios, a docker-compose example is available in examples/docker-compose.yaml. This demonstrates how to run the MCP server with Docker Compose, including resource limits and security options.
Note: Most MCP clients manage container lifecycle directly, so docker-compose is primarily useful for testing or custom deployment scenarios.
Debugging
Since MCP servers communicate over stdio, debugging can be challenging. We recommend using the MCP Inspector, which is available as a package script:
npm run inspectorThe Inspector will provide a URL to access debugging tools in your browser.
Contributing
Please see CONTRIBUTING.md for development setup and guidelines.
License
MIT License - see the LICENSE file for details
Support
- Documentation: docs.glean.com
- Issues: GitHub Issues
