Codebase Indexing MCP Server

What it does

This MCP server enables AI assistants to perform deep semantic search across a local or remote codebase. It parses code into meaningful chunks using tree-sitter, generates embeddings via Voyage AI, and stores them in Qdrant for lightning-fast similarity retrieval. This allows the AI to find relevant code based on meaning rather than just keyword matches.

Tools

• index_directory: Recursively indexes all files in a specified directory.
• codebase_search: Natural language search for high-level codebase understanding.
• search_code: Semantic similarity search for specific code chunks.
• search_functions: Find functions by name or description.
• search_classes: Find classes by name or description.
• ingest_git_repository: Clones and indexes a Git repo remotely.
• get_code_context: Retrieves context around a specific code chunk.
• get_indexing_stats: Provides health and volume metrics of the index.

Installation

Add the following to your claude_desktop_config.json:

{
  "mcpServers": {
    "codebase-indexing": {
      "command": "npx",
      "args": ["mcp-codebase-indexing-server"],
      "env": {
        "VOYAGE_API_KEY": "your_voyage_api_key",
        "QDRANT_URL": "http://localhost:6333"
      }
    }
  }
}

Supported hosts

• Cursor
• Claude Desktop