Getting Started

Get Glaux file intelligence into your AI coding agent in under a minute. No install, no API keys — just connect to the cloud MCP server.

Prerequisites

Your repository must be seeded before Glaux tools return useful data. Showcase repos (vercel/next.js, tobi/qmd, nousresearch/hermes-agent) are already seeded and ready to use. For your own repo, request it here — we'll seed it and keep it updated for free.

Connect your agent

Claude Code

One command:

claude mcp add glaux --transport http https://www.glaux.dev/api/mcp

Cursor

Add to .cursor/mcp.json in your project root:

{
  "mcpServers": {
    "glaux": {
      "url": "https://www.glaux.dev/api/mcp"
    }
  }
}

Windsurf

Add to your MCP configuration:

{
  "mcpServers": {
    "glaux": {
      "serverUrl": "https://www.glaux.dev/api/mcp"
    }
  }
}

Any MCP-compatible agent

Point your agent's MCP config at:

https://www.glaux.dev/api/mcp

The transport is HTTP — no local process, no dependencies. Your agent connects directly to the Glaux cloud.

What your agent gets

Once connected to a seeded repo, the agent gets two tools:

• glaux_file_intel — call before editing any file. Returns change risk, hidden coupling, files to review first, and volatility.
• glaux_pr_review — call before reviewing or merging a PR. Returns per-file risk and coupling signals across all changed files.

See MCP Tools for the full tool reference. To have your agent call these tools automatically before every edit, see the Agent Skill.

Alternative: REST API

If your tooling doesn't support MCP, use the REST API directly. All showcase repos are publicly accessible — no auth required.

# Get file intelligence
curl https://www.glaux.dev/api/repos/tobi/qmd/file/src/llm.ts

# Get PR context
curl https://www.glaux.dev/api/repos/tobi/qmd/pr/10

# Get repo overview
curl https://www.glaux.dev/api/repos/tobi/qmd

What gets indexed?

Glaux analyzes your repository to compute:

• Churn metrics — commit count, additions/deletions, recent activity
• Co-change patterns — files that historically change together
• Importance & connectedness — which files are central in the co-change network
• Cluster detection — file groups identified via Leiden algorithm
• Momentum — whether co-change coupling is increasing, decreasing, or stable
• Issue links — GitHub issues and PRs linked to each file