codegraph

Graph RAG for code — semantic search with relationship-aware context for coding agents.

Early Research Phase

Semantic Search

Find code by meaning, not just text matching. Hybrid search combines keyword matching (FTS5) with graph traversal for relationship-aware results.

Relationship-Aware Context

When you find a function, also see what calls it, what it calls, its parent class, and file imports — all in one query.

Always Fresh

File watcher monitors your codebase and incrementally re-indexes changed files. Respects .gitignore and debounces rapid changes.

Agent-Friendly

Built as an MCP server for AI coding agents like Claude Code. Five tools give agents structural understanding of any codebase.

How It Works

Codegraph parses your source code with tree-sitter, extracts symbols and their relationships, and stores the resulting graph in SQLite. An MCP server exposes this graph to AI coding agents through structured search and navigation tools.

Source Code → Tree-sitter Parser → Symbol Graph → SQLite → MCP Server → AI Agent

Supported Languages

Language	Extensions
Go	`.go`
TypeScript	`.ts`, `.tsx`
JavaScript	`.js`, `.jsx`
Python	`.py`

Quick Start

# Install
go install github.com/skaramicke/codegraph@latest

# Index your project
codegraph index --root /path/to/project

# Search from the command line
codegraph search "handleAuth"

# Or run as an MCP server for Claude Code
codegraph serve

See the Getting Started guide for full setup instructions.