CodeGraph
CodeGraph is a local-first code intelligence library, CLI, and MCP server that gives AI coding agents a pre-indexed knowledge graph of a codebase.
Overview
CodeGraph helps AI coding agents understand repositories faster by indexing code into a local knowledge graph, exposing symbols, relationships, imports, files, and call structures through MCP instead of relying on repeated file scanning.
Core Features & Capabilities
Ideal for software developers, AI coding agent users, Claude Code users, Codex CLI users, Cursor users, OpenCode users, Gemini CLI users, technical founders, engineering teams, open-source maintainers, DevOps engineers, codebase refactoring teams, AI tool builders, and developers working with medium to large repositories.
- Index a codebase locally so AI coding agents can query structure instead of repeatedly scanning files
- Expose symbols, imports, relationships, files, and call structures through an MCP server
- Use deterministic tree-sitter parsing rather than LLM-generated summaries for code understanding
- Reduce token usage and tool calls for architecture questions, refactors, debugging, and repository exploration
- Work with Claude Code, Codex CLI, Cursor, OpenCode, Gemini CLI, Kiro, Antigravity IDE, and other agentic coding tools
Trending Use Cases
Why Developers Choose CodeGraph
Visit the CodeGraph GitHub repository, install the npm package, and run the CodeGraph setup for your project. Index the repository so CodeGraph can parse supported files, store the generated symbols and relationships in the local .codegraph directory, and start the MCP server for your preferred coding agent. Once connected, use Claude Code, Codex CLI, Cursor, OpenCode, Gemini CLI, or another compatible agent to ask architecture, symbol, dependency, refactoring, and debugging questions with access to the pre-indexed local code graph.
“CodeGraph gives AI coding agents a local pre-indexed map of your codebase so they can understand structure faster with fewer tokens and fewer tool calls.”
Getting Started with CodeGraph
By combining local-first indexing, tree-sitter parsing, SQLite storage, full-text search, MCP access, symbol relationships, file structure, imports, and agent-compatible tooling, CodeGraph gives developers a practical way to make AI coding agents more efficient when working with real codebases.
Open the tool and review its core product experience.
Create your account or access your existing workspace.
Use your own task to judge speed, quality, and fit.
Check similar AI tools before making a final decision.
CodeGraph
Colby McHenry is the creator of CodeGraph, an open-source local-first code intelligence library, CLI, and MCP server for AI coding agents, focused on code knowledge graphs, tree-sitter parsing, SQLite indexing, and token-efficient repository understanding.
CodeGraph is an open-source developer tool designed to help AI coding agents understand codebases faster and more efficiently. Instead of forcing agents to repeatedly scan files with grep, glob, and read operations, CodeGraph builds a local pre-indexed knowledge graph from the project’s source code. It parses supported codebases using tree-sitter, stores symbols, files, and relationships in SQLite with full-text search, and exposes codebase intelligence through an MCP server. This allows tools such as Claude Code, Codex CLI, Cursor, OpenCode, Gemini CLI, Kiro, Antigravity IDE, and Hermes Agent to query symbols, call relationships, imports, files, and code structure with fewer tokens and fewer tool calls. Because CodeGraph is local-first and deterministic, it keeps project data on the user’s machine and derives knowledge from the code’s syntax tree rather than LLM-generated summaries.
github.com/colbymchenry/codegraph
Comments (0)
No Comments Found