Install guide and config at curatedmcp.com

mcp-repo-graph: Stop Grepping, Start Mapping

Every developer knows the feeling: you're dropped into a 10,000-file codebase and need to fix a bug. You grep for function names. You read imports. You jump between files. Your AI assistant does the same thing—reading entire modules just to find the three files that actually matter.

mcp-repo-graph changes that. It's an MCP server that gives AI coding assistants a structural map of your codebase before they start digging. Instead of blind file-hunting, they query a graph of entities, relationships, and feature flows to pinpoint exactly what needs attention.

What It Does

repo-graph scans your repository once and builds a lightweight dependency graph: modules, classes, functions, routes, services, components, and how they connect (imports, HTTP calls, cross-stack flows). The graph is queryable through 11 MCP tools—flow, trace, impact, find, status, and others—that let your AI assistant answer questions like:

• "Which files handle user authentication?"
• "What breaks if I change this function?"
• "Where does this API route connect to the backend?"

It works across 20+ languages (Go, Rust, TypeScript, Python, Java, C#, and more) and understands cross-stack relationships between frontend calls and backend routes. The scanning engine is written in native Rust and shipped as a prebuilt wheel, so even large repos map in seconds.

In a controlled test on a Go+Angular codebase, the same task with the same model used 2.5× fewer tokens and ran ~9× faster with repo-graph enabled.

How to Install

uvx mcp-repo-graph --repo /path/to/your/project

Add to your claude_desktop_config.json:

{
  "mcpServers": {
    "mcp-repo-graph": {
      "command": "uvx mcp-repo-graph --repo /path/to/your/project"
    }
  }
}

Real-World Use Cases

• Bug triage in unfamiliar code. Paste an error message. Your assistant queries the graph to find which module threw it, traces callers, and pulls only the relevant files. No reading half the codebase.
• Feature impact analysis. Refactor a utility function. Before touching it, your assistant traces dependents and surfaces all affected routes, components, and tests.
• Onboarding faster. New to the team? Ask Claude to map out how a feature flows end-to-end: from frontend button click → API call → backend handler → database write. The graph answers in milliseconds.

Full install guides for Claude Desktop, Cursor, Windsurf, and more at CuratedMCP.