π Open source Claude Code CLI source code. Advanced AI Agent for developers. Includes TypeScript codebase for LLM tool-calling, agentic workflows, and terminal UI. Remember this is just the skeleton not the brain itself. Found by Chaofan Shou.
# Add to your Claude Code skills
git clone https://github.com/yasasbanukaofficial/claude-codePS: This breakdown is also available on this blog with a better reading experience and UX :)
Note: There's a non-zero chance this repo might be taken down. If you want to play around with it later or archive it yourself, feel free to fork it and bookmark the external blog link!
I did not leak these files. I have simply provided an easy, documented way to access and study this codebase for research purposes. All files and information originate from public findings shared on Twitter/X. All credit for the discovery goes to the original source.
Earlier today (March 31st, 2026) - Chaofan Shou (@Fried_rice) discovered something that Anthropic probably didn't want the world to see: the entire source code of Claude Code, Anthropic's official AI coding CLI, was sitting in plain sight on the npm registry via a sourcemap file bundled into the published package.
This repository is a backup of that leaked source, providing a full breakdown of what's in it, how the leak happened, and the internal systems that were never meant to be public.
When you publish a JavaScript/TypeScript package to npm, the build toolchain often generates source map files (.map files). These files bridge minified production code and the original source for debugging.
No comments yet. Be the first to share your thoughts!
The catch? Source maps contain the original source code embedded as strings inside a JSON file under the sourcesContent key.
{
"version": 3,
"sources": ["../src/main.tsx", "../src/tools/BashTool.ts", "..."],
"sourcesContent": ["// The ENTIRE original source code of each file", "..."],
"mappings": "AAAA,SAAS,OAAO..."
}
By forgetting to add *.map to .npmignore or failing to disable source maps in production builds (Bun's default behavior), the entire raw source was shipped to the npm registry.
Claude Code is not just a simple CLI. It's a massive 785KB main.tsx entry point featuring a custom React terminal renderer (Ink), 40+ tools, and complex multi-agent orchestration.
Inside src/buddy/, there is a full Tamagotchi-style companion system.
userId.DEBUGGING, CHAOS, and SNARK, with a "soul" description written by Claude.Anthropic employees use Claude Code to contribute to public repos. Undercover Mode (src/utils/undercover.ts) prevents the AI from leaking internal info:
Claude Code "dreams" to consolidate memory. The autoDream service (src/services/autoDream/) runs as a background subagent to:
MEMORY.md.src/
βββ main.tsx # CLI Entrypoint (Commander.js + React/Ink)
βββ QueryEngine.ts # Core LLM logic (~46K lines)
βββ Tool.ts # Base tool definitions
βββ tools/ # 40+ Agent tools (Bash, Files, LSP, Web)
βββ services/ # Backend (MCP, OAuth, Analytics, Dreams)
βββ coordinator/ # Multi-agent orchestration (Swarm)
βββ bridge/ # IDE Integration layer
βββ buddy/ # The secret Tamagotchi system
Clone the repository:
git clone https://github.com/your-username/claude-leaked.git
cd claude-leaked
Install Dependencies:
npm install
Build the Project:
npm run build
Run the CLI:
node dist/main.js
This repo includes an MCP Server to let you explore the source using Claude itself:
claude mcp add code-explorer -- npx -y claude-code-explorer-mcp
Keywords: Claude Code Leak, Anthropic Source Code, AI Agent Framework, Claude 3.5 Sonnet CLI, Tengu Anthropic, npm sourcemap leak, Open Source AI Agent.
Disclaimer: All original source code is the proprietary property of Anthropic PBC. This repository is for educational and archival purposes only. This is not an official Anthropic product.
For spamming reasons the email has been removed.