CodeGraph

CodeGraph builds a local code knowledge graph for AI coding agents (Claude Code, Cursor, Codex CLI, OpenCode), enabling a single tool call to retrieve entry points, related symbols, and code snippets without expensive exploration scans. Benchmarks show an average 92% reduction in tool calls and 71% faster exploration.

Overview

The project uses tree-sitter to parse source code into ASTs, extracting nodes (functions, classes, methods) and edges (calls, imports, extends, implements) via language-specific queries. Everything is stored in a local SQLite database (.codegraph/codegraph.db) with FTS5 full-text search. The MCP server watches for file changes using native OS events and auto-syncs incrementally.

Supports 19+ programming languages and framework-aware route recognition for 13 web frameworks including Django, Flask, FastAPI, Express, Laravel, Rails, Spring, Gin, and more.

Key Features

• Smart context building: One tool call returns entry points, related symbols, and code snippets
• Full-text search: Instant code name search powered by SQLite FTS5
• Impact analysis: Trace callers, callees, and full impact radius of any symbol
• Always fresh: File watcher uses native OS events with debounced incremental sync, zero config
• 19+ language support: TypeScript, JavaScript, Python, Go, Rust, Java, C#, PHP, Ruby, C/C++, Swift, Kotlin, Dart, and more
• Framework-aware routes: Recognizes routing files for 13 web frameworks and links URL patterns to handlers
• 100% local: No data leaves your machine, no API keys, no external services
• Multi-agent support: Claude Code, Cursor, Codex CLI, OpenCode

Use Cases

• Large codebase navigation: Quickly locate entry points and call chains in million-line projects
• Change impact assessment: Trace full impact radius before making code changes to prevent regressions
• AI agent acceleration: Reduce agent exploration rounds, lower token consumption and latency
• Code review assistance: Rapidly understand code structure and dependencies involved in a PR

Technical Highlights

• Language: TypeScript
• Storage: Local SQLite + FTS5 full-text search
• Parsing engine: tree-sitter multi-language AST parsing
• Protocol: MCP (Model Context Protocol)
• Node.js requirement: >= 18.0.0
• MCP tools: codegraph_search, codegraph_context, codegraph_callers, codegraph_callees, codegraph_impact, and 3 more
• License: MIT