What is larkx?

larkx is an AI codebase indexer and MCP server. Instead of reading raw source files, your AI agent queries a compact graph - same understanding, far fewer tokens.

Works with Claude Code, Cursor, GitHub Copilot, Gemini CLI, and OpenAI Codex.

Install

npm install -g larkx

Quick start

cd your-project
larkx init      # one-time setup: MCP, agent files, hooks
larkx index     # build the index

larkx init creates instruction files for each agent you use (CLAUDE.md, .cursorrules, AGENTS.md, GEMINI.md). These files are auto-updated on every larkx index run.

Supported agents

| Agent | MCP | Context file | |-------|:---:|:------------:| | Claude Code | ✓ | ✓ | | Cursor | ✓ | ✓ | | GitHub Copilot | - | ✓ | | OpenAI Codex | - | ✓ | | Gemini CLI | - | ✓ |

Supported languages

Current indexing targets these extensions:

• JavaScript / TypeScript: .js, .jsx, .ts, .tsx, .mjs, .cjs
• Python: .py
• Go: .go
• Rust: .rs
• Java: .java
• C/C++: .c, .cpp
• C#: .cs

Files in other languages are currently skipped.

CLI reference

| Command | What it does | |---------|-------------| | larkx init | Setup wizard: MCP, agent instruction files, hooks | | larkx index | Build or update the index; auto-refreshes agent files | | larkx index --ai | Add AI summaries per file | | larkx index --watch | Keep the index live as you edit | | larkx bench | Token benchmark — measure actual savings on your project | | larkx stats | Token estimates per level | | larkx context | Print the index to stdout | | larkx search <name> | Find a function or class by name | | larkx impact <file> | List every file that imports a given file | | larkx deadcode | Find unreachable files and functions | | larkx serve | Open the visual graph in your browser | | larkx mcp --check | Health-check the MCP server |

Documentation

Full setup guides, MCP integration, token optimization, and FAQ at Documentation.