CodeGenome
BetaOgro Projukti · Developer Tools & AI
Turn your codebase into a living knowledge graph.
- GitHub stars
- 23
- License
- MIT

>About CodeGenome
Standard context tools dump flat, truncated text into an LLM's window, causing token bloat and architectural hallucinations. CodeGenome treats code like a living connectome instead.
It runs localized, incremental tree-sitter passes and tracks changes in SQLite, building a knowledge graph of files, symbols, imports and calls in real time — without dragging down your machine.
As your AI agent writes and refactors code, CodeGenome maps the new relationships instantly and serves them to Cursor, Claude Desktop or any MCP client, so the agent reasons about your whole architecture. Run it locally, on servers or headless in CI.
- Version
- 0.1.4
- Updated
- 1 Jun 2026
- Available for
- Python 3.11+, CLI / TUI, MCP (HTTP & stdio)
- Languages
- English
- Price
- Free
- License
- MIT
>Screenshots
>Key features
Incremental knowledge graph
Parses source with tree-sitter and tracks changes in SQLite, keeping a structural map of files, symbols, imports and calls.
MCP server for AI agents
Serves high-fidelity context to Cursor, Claude, Copilot and custom clients over HTTP or stdio.
Live graph & watch mode
Watch the graph evolve as you or your agent code — and share it with other devices on your LAN.
Architecture insights
Filter for god classes, fragile nodes, tech debt and bridge nodes to find the parts of a system that need attention.
Terminal UI
Analyse, export, start the MCP server and run live mode from a keyboard-driven TUI.
Exports
JSON, HTML & Markdown, GraphML, Cypher for Neo4j, and Obsidian vaults.
Multi-language parsing
Python, JavaScript, TypeScript, Go and Rust grammars out of the box.
- GitHub Stars
- 23
- Forks
- 8
- PyPI Downloads
- 737
>How it works
- 01
Analyse
codegenome analyze parses your workspace with tree-sitter into a local graph.
- 02
Track
Incremental passes keep the graph in sync with every change, stored in SQLite under .genome/.
- 03
Serve
The MCP server exposes the graph to your editor's AI agent as precise context.
- 04
Explore
Use the live graph, the TUI or exports to understand your architecture.
>Get started
Requirements
- Python 3.11 or newer
- pip
pip install codegenomecd /path/to/your/projectcodegenome analyze .codegenome mcp-startcodegenome evolve --live .codegenome export --format obsidian --path .codegenome tui>Changelog
v0.1.4
- Share the live graph with other devices on your LAN (evolve --live --lan)
v0.1.0
- First release on PyPI
>Ratings & reviews
—
0 reviews
- 5 ★0% (0)
- 4 ★0% (0)
- 3 ★0% (0)
- 2 ★0% (0)
- 1 ★0% (0)