Repo GraphRAG

Provides an MCP server that builds, updates, and queries a knowledge graph from code and text documents for implementation planning and Q&A.

python

GitHub Stars

python

Language

6 months ago

First Indexed

2 months ago

Catalog Refreshed

Documentation & install

Readme and setup notes from the catalogue, plus a client-ready config you can copy for your MCP host.

Installation

Add the following to your MCP client configuration file.

Configuration

View docs

{
  "mcpServers": {
    "yumeiriowl-repo-graphrag-mcp": {
      "command": "uv",
      "args": [
        "--directory",
        "/absolute/path/to/repo-graphrag-mcp",
        "run",
        "server.py"
      ],
      "env": {
        "GRAPH_CREATE_PROVIDER": "anthropic",
        "GRAPH_ANALYSIS_PROVIDER": "anthropic",
        "GRAPH_CREATE_MODEL_NAME": "claude-3-5-haiku-20241022",
        "GRAPH_ANALYSIS_MODEL_NAME": "claude-sonnet-4-20250514"
      }
    }
  }
}

You set up and run a dedicated MCP server that builds a knowledge graph from code and text in a repository, then uses that graph for planning and Q&A. This enables you to plan concrete implementation steps and answer questions based on the built graph, with support for incremental updates and optional vector search.

How to use

Launch and connect to the MCP server from a client, then use the graph tools to create, plan, and query the knowledge graph. You will build the graph from your repository, generate a planning strategy for requested changes, and ask questions guided by the graph.

Key tools you will use via the MCP client are graph_create, graph_plan, and graph_query. graph_create analyzes a directory to build or update the knowledge graph and its embedding index. graph_plan generates concrete implementation steps for a requested change. graph_query answers questions based on the knowledge graph, optionally combined with vector search.

Below is a sample client configuration for starting the MCP server locally using the standard VS Code Copilot-style MCP client configuration. This demonstrates a local, stdio-based integration you can reproduce in your environment.

{
  "servers": {
    "repo_graphrag": {
      "type": "stdio",
      "command": "uv",
      "args": [
        "--directory",
        "/absolute/path/to/repo-graphrag-mcp",
        "run",
        "server.py"
      ]
    }
  }
}

Available tools

graph_create

Build or update the knowledge graph from a target repository/directory and generate the embedding index. Supports incremental updates and optional vector search integration.

graph_plan

Provide a detailed implementation plan and concrete change steps based on the knowledge graph, optionally combined with vector search.

graph_query

Answer questions about the target repository/directory using the knowledge graph, optionally combined with vector search.

Built by

VeilStrat

AI signals for GTM teams