code-graph-mcp

A high-performance code knowledge graph server implementing the Model Context Protocol (MCP). Indexes codebases into a structured AST knowledge graph with semantic search, call graph traversal, and HTTP route tracing — designed to give AI coding assistants deep, structured understanding of your code.

Features

• Multi-language parsing — Tree-sitter based AST extraction for TypeScript, JavaScript, Go, Python, Rust, Java, C, C++, HTML, CSS
• Semantic code search — Hybrid BM25 full-text + vector semantic search with Reciprocal Rank Fusion (RRF), powered by sqlite-vec
• Call graph traversal — Recursive CTE queries to trace callers/callees with cycle detection
• HTTP route tracing — Map route paths to backend handler functions (Express, Flask/FastAPI, Go)
• Incremental indexing — Merkle tree change detection with file system watcher for real-time updates
• Context compression — Token-aware snippet extraction for LLM context windows
• Embedding model — Optional local embedding via Candle (feature-gated embed-model)
• MCP protocol — JSON-RPC 2.0 over stdio, plug-and-play with Claude Code, Cursor, and other MCP clients

Architecture

src/
├── mcp/          # MCP protocol layer (JSON-RPC, tool registry, server)
├── parser/       # Tree-sitter parsing, relation extraction, language support
├── indexer/      # 3-phase pipeline, Merkle tree, file watcher
├── storage/      # SQLite schema, CRUD, FTS5 full-text search
├── graph/        # Recursive CTE call graph queries
├── search/       # RRF fusion search combining BM25 + vector
├── embedding/    # Candle embedding model (optional)
├── sandbox/      # Context compressor with token estimation
└── utils/        # Language detection, config

Quick Start

Prerequisites

• Rust 1.75+ (2021 edition)
• A C compiler (for bundled SQLite / sqlite-vec)

Build

# Default build (with local embedding model)
cargo build --release

# Without embedding model (lighter build)
cargo build --release --no-default-features

Run

The server communicates via JSON-RPC 2.0 over stdin/stdout:

cd /path/to/your/project
code-graph-mcp

Configure with Claude Code

Add to your MCP settings:

{
  "mcpServers": {
    "code-graph": {
      "command": "/path/to/code-graph-mcp",
      "cwd": "/path/to/your/project"
    }
  }
}

MCP Tools

| Tool | Description | |------|-------------| | semantic_code_search | Hybrid BM25 + vector + graph search for AST nodes | | get_call_graph | Trace upstream/downstream call chains for a function | | find_http_route | Map route path to backend handler function | | get_ast_node | Extract a specific code symbol from a file | | read_snippet | Read original code snippet by node ID with context | | start_watch / stop_watch | Start/stop file system watcher for incremental indexing | | get_index_status | Query index status and health | | rebuild_index | Force full index rebuild |

Supported Languages

TypeScript, JavaScript, Go, Python, Rust, Java, C, C++, HTML, CSS

Storage

Uses SQLite with:

• FTS5 for full-text search
• sqlite-vec extension for vector similarity search
• Merkle tree hashes for incremental change detection

Data is stored in .code-graph/index.db under the project root (auto-created, gitignored).

Development

# Run tests
cargo test

# Run tests without embedding model
cargo test --no-default-features

# Check compilation
cargo check

License

See LICENSE for details.