GraphRAG

GraphRAG provides a knowledge graph of your codebase — capturing classes, functions, types, and their relationships — so you can ask architectural questions that go beyond text similarity search.

How it works

RagGo uses a NetworkX-based graph built from a pre-generated API_graph.pkl file. The graph captures:

• Classes, structs, interfaces, and their inheritance hierarchies
• Functions, methods, and their call relationships
• Module and package structure

You query the graph through MCP tools to explore relationships, find callers, and understand architecture at a high level.

MCP tools for GraphRAG

GraphRAG is accessed entirely through MCP tools from Claude, Cursor, or any connected client:

// Search API documentation graph
{
  "tool": "search_api_docs",
  "input": {
    "query": "authentication"
  }
}

// Get detailed class information
{
  "tool": "get_class_info",
  "input": {
    "class_name": "EmbedService"
  }
}

// Get class hierarchy
{
  "tool": "get_class_hierarchy",
  "input": {
    "class_name": "Server"
  }
}

// Get function details
{
  "tool": "get_function_details",
  "input": {
    "function_name": "ParseDocument"
  }
}

// Explore graph statistics
{
  "tool": "explore_graph_stats",
  "input": {}
}

Supported languages

• Go (full support)
• Python (class/function extraction)
• TypeScript / JavaScript (class, function, import extraction)
• Rust (struct, impl, trait extraction)