glean-mcp 3.3.0

Python client + MCP server adapter for Glean (search, chat, read documents). Provides a reusable HTTP client plus simple sync helper functions.

Glean MCP Server & Python Package

Simple, focused implementation providing:

• Cookie-based client (glean_mcp.cookie_client.GleanClient) — uses browser session cookies
• Token-based client (glean_mcp.token_client.TokenBasedGleanClient) — uses API tokens for server-to-server auth
• MCP server (python -m glean_mcp.server) — auto-detects authentication method; ready for Docker

Links: Releases · Cookie Guide · Token Auth Guide

Install

pip install -U glean-mcp

Package Layout

src/
└── glean_mcp/
    ├── cookie_client.py      # Cookie-based authentication client
    ├── token_client.py       # Token-based authentication client
    ├── server.py             # MCP server with auto-detection
    └── glean_filter.py       # Response filtering utilities

Quick Start

MCP Server (local or Docker)

The MCP server automatically detects your authentication method.

# Choose ONE (token preferred)
export GLEAN_API_TOKEN="your-api-token"   # preferred
# OR
export GLEAN_COOKIES="your-browser-cookies"

# Set your Glean instance
export GLEAN_BASE_URL="https://your-company-be.glean.com"  # or set GLEAN_INSTANCE=your-company

# Run the MCP server locally
python -m glean_mcp.server

Docker (example):

docker run --pull always --rm \
  -e GLEAN_API_TOKEN="$GLEAN_API_TOKEN" \
  -e GLEAN_BASE_URL="$GLEAN_BASE_URL" \
  ghcr.io/alankyshum/glean-mcp-server:latest

Upgrade: use --pull always (Docker) or pip install -U glean-mcp (pip).

Library Usage (async)

Cookie-based:

from glean_mcp import GleanClient

client = GleanClient(base_url, cookies)
results = await client.search("onboarding docs")
await client.close()

Token-based:

from glean_mcp import TokenBasedGleanClient

client = TokenBasedGleanClient(base_url, api_token)
results = await client.search("onboarding docs")
await client.close()

Auto-detection (same logic as server):

from glean_mcp import create_glean_client

client = create_glean_client()  # Uses env vars; prefers token
results = await client.search("onboarding docs")
await client.close()

Authentication

Two supported methods:

🍪 Cookies (original)

• Use browser cookies from your Glean session
• Requires GLEAN_COOKIES
• Uses /api/v1/ endpoints
• See Cookie Guide

🔑 Token (recommended)

• Use Glean API tokens for server-to-server authentication
• Requires GLEAN_API_TOKEN
• Uses /rest/api/v1/ endpoints
• More secure for automated/production environments
• See Token Auth Guide

🤖 Auto-detection rules

1. If GLEAN_API_TOKEN is set → token-based auth
2. Else if GLEAN_COOKIES is set → cookie-based auth
3. If both are set → token preferred
4. If neither is set → error with guidance

Environment Variables

Required for server/library:

• GLEAN_BASE_URL (e.g. https://your-company-be.glean.com) or GLEAN_INSTANCE
• One of: GLEAN_API_TOKEN (preferred) or GLEAN_COOKIES

Optional (server behavior):

• GLEAN_DEFAULT_PAGE_SIZE (default: 14)
• GLEAN_DEFAULT_SNIPPET_SIZE (default: 215)
• GLEAN_TOOL_DESCRIPTION (tool description text)
• GLEAN_AUTO_OPEN_BROWSER (default: true)

Development

git clone https://github.com/alankyshum/glean-mcp-server.git
cd glean-mcp-server
python -m venv .venv && source .venv/bin/activate
pip install -e '.[dev]'
pytest -q

Versioning & Publishing

• Semantic versioning; breaking changes bump MAJOR
• Tag vX.Y.Z after updating version strings; CI publishes to PyPI/GHCR if versions match

License

MIT