Claude Code Source Snapshot

An exploratory mirror of a source snapshot reportedly exposed via published source maps on March 31, 2026

The raw imported snapshot is preserved in this repository's backup branch. The main branch contains added documentation, tooling, and repository metadata.

The Story: How the Source Code Leaked

On March 31, 2026, the developer community was stunned when the entire uncompiled TypeScript source code for Anthropic's flagship CLI tool, Claude Code, was exposed to the public. The leak wasn't the result of a sophisticated hack or an insider threat, but a simple build configuration oversight.

Security researcher Chaofan Shou (@Fried_rice) discovered that the officially published npm registry package for Claude Code accidentally included a .map (source map) file. This file contained a direct reference to Anthropic's R2 storage bucket, allowing anyone to download a zip archive containing the full, unobfuscated TypeScript source code of the project.

"Claude code source code has been leaked via a map file in their npm registry!"

— @Fried_rice, March 31, 2026

The leaked code revealed a massive, highly sophisticated terminal application containing over 512,000 lines of code across roughly 1,900 files. It gave developers an unprecedented look into how one of the world's leading AI labs builds internal tooling, Agent Swarms, and integrations with the Model Context Protocol (MCP).

Full Breakdown of Code Findings

By analyzing the leaked repository, developers uncovered several fascinating architectural decisions and advanced systems powering Claude Code:

1. The Terminal is a React App

Perhaps the most surprising finding is that the entire CLI is built using React and Ink. Instead of traditional stdout printing, Claude Code functions identically to a modern React web app with hooks, contexts, and over 140 UI components—all rendering directly into the terminal interface using chalk for styling.

2. Bun Runtime & Dead Code Elimination

Instead of Node.js, Claude Code relies on the Bun runtime. This allows it to run TSX files natively without a separate transpilation step. The codebase makes heavy use of Bun's bun:bundle feature for dead-code elimination, stripping out unreleased features via flags like PROACTIVE (autonomous agent mode), WORKFLOW_SCRIPTS, and VOICE_MODE.

3. Agent Swarms and Multi-Agent Architecture

The code includes an advanced QueryEngine.ts (a massive ~46K line file) that handles LLM tool-calling loops. More importantly, it features an AgentTool and a coordinator/ subsystem that allows Claude to spawn sub-agents to parallelize tasks. Agents can even communicate with each other using a SendMessageTool and form groups via a TeamCreateTool.

4. Comprehensive Tool & Command Ecosystem

Claude Code features around 40 advanced tools ranging from standard shell execution (BashTool) and file editing (FileEditTool), to powerful search modules (GrepTool powered by ripgrep). It also contains a robust command system with ~50 slash commands (like /commit, /review, /mcp), some of which return full React JSX components directly to the user's terminal to render diagnostics or UI screens.

5. IDE "Bridge" System

A bridge/ subsystem acts as a bidirectional communication layer, securely connecting the CLI tool with IDE extensions (like VS Code and JetBrains) using JWT authentication, allowing the terminal agent to drive the editor and read its state.

What Is Claude Code?

Claude Code is Anthropic's official CLI tool for interacting with Claude directly from the terminal. This leaked source code repository acts as a breakthrough artifact showing how modern LLM interfaces work: editing files, running commands, searching codebases, and managing git workflows using Model Context Protocol (MCP) and multi-agent systems.

This repository contains the complete source snapshot alongside newly added docs, MCP tooling, and repository metadata to help inspect it. Discover the power of agent swarms, terminal voice modes, and IDE bridges hidden deep in the React & Ink architecture!


Leaked Artifact	2026-03-31
Language	TypeScript (strict)
Runtime Architecture	Bun
Terminal UI Framework	React + Ink
Code Scale	~1,900 files · 512,000+ lines of codebase

Documentation

For in-depth guides, see the docs/ directory:

Guide	Description
Architecture	Core pipeline, startup sequence, state management, rendering, data flow
Tools Reference	Complete catalog of all ~40 agent tools with categories and permission model
Commands Reference	All ~85 slash commands organized by category
Subsystems Guide	Deep dives into Bridge, MCP, Permissions, Plugins, Skills, Tasks, Memory, Voice
Exploration Guide	How to navigate the codebase — study paths, grep patterns, key files

Also see: CONTRIBUTING.md · MCP Server README

Explore with MCP Server

This repo also ships an MCP server that lets any MCP-compatible client (Claude Code, Claude Desktop, VS Code Copilot, Cursor) explore the snapshot interactively.

Install from npm

The MCP server is published as claude-code-explorer-mcp on npm — no need to clone the repo:

# Claude Code
claude mcp add claude-code-explorer -- npx -y claude-code-explorer-mcp

One-liner setup (from source)

git clone https://github.com/muhammadhaseebiqbal-dev/claude-code-leak-source-code.git ~/claude-code-source-code \
  && cd ~/claude-code-source-code/mcp-server \
  && npm install && npm run build \
  && claude mcp add claude-code-explorer -- node ~/claude-code-source-code/mcp-server/dist/index.js

Step-by-step setup

# 1. Clone the repo
git clone https://github.com/muhammadhaseebiqbal-dev/claude-code-leak-source-code.git
cd claude-code-source-code/mcp-server

# 2. Install & build
npm install && npm run build

# 3. Register with Claude Code
claude mcp add claude-code-explorer -- node /absolute/path/to/claude-code-source-code/mcp-server/dist/index.js

Replace /absolute/path/to/claude-code-source-code with your actual clone path.

VS Code / Cursor / Claude Desktop config

VS Code — add to .vscode/mcp.json:

{
  "servers": {
    "claude-code-explorer": {
      "type": "stdio",
      "command": "node",
      "args": ["${workspaceFolder}/mcp-server/dist/index.js"],
      "env": { "CLAUDE_CODE_SRC_ROOT": "${workspaceFolder}/src" }
    }
  }
}

Claude Desktop — add to your config file:

{
  "mcpServers": {
    "claude-code-explorer": {
      "command": "node",
      "args": ["/absolute/path/to/claude-code-source-code/mcp-server/dist/index.js"],
      "env": { "CLAUDE_CODE_SRC_ROOT": "/absolute/path/to/claude-code-source-code/src" }
    }
  }
}

Cursor — add to ~/.cursor/mcp.json (same format as Claude Desktop).

Available tools & prompts

Tool	Description
`list_tools`	List all ~40 agent tools with source files
`list_commands`	List all ~50 slash commands with source files
`get_tool_source`	Read full source of any tool (e.g. BashTool, FileEditTool)
`get_command_source`	Read source of any slash command (e.g. review, mcp)
`read_source_file`	Read any file from `src/` by path
`search_source`	Grep across the entire source tree
`list_directory`	Browse `src/` directories
`get_architecture`	High-level architecture overview

Prompt	Description
`explain_tool`	Deep-dive into how a specific tool works
`explain_command`	Understand a slash command's implementation
`architecture_overview`	Guided tour of the full architecture
`how_does_it_work`	Explain any subsystem (permissions, MCP, bridge, etc.)
`compare_tools`	Side-by-side comparison of two tools

Try asking: "How does the BashTool work?" · "Search for where permissions are checked" · "Show me the /review command source"

Custom source path / Remove

# Custom source location
claude mcp add claude-code-explorer -e CLAUDE_CODE_SRC_ROOT=/path/to/src -- node /path/to/mcp-server/dist/index.js

# Remove
claude mcp remove claude-code-explorer

Directory Structure

src/
├── main.tsx                 # Entrypoint — Commander.js CLI parser + React/Ink renderer
├── QueryEngine.ts           # Core LLM API caller (~46K lines)
├── Tool.ts                  # Tool type definitions (~29K lines)
├── commands.ts              # Command registry (~25K lines)
├── tools.ts                 # Tool registry
├── context.ts               # System/user context collection
├── cost-tracker.ts          # Token cost tracking
│
├── tools/                   # Agent tool implementations (~40)
├── commands/                # Slash command implementations (~50)
├── components/              # Ink UI components (~140)
├── services/                # External service integrations
├── hooks/                   # React hooks (incl. permission checks)
├── types/                   # TypeScript type definitions
├── utils/                   # Utility functions
├── screens/                 # Full-screen UIs (Doctor, REPL, Resume)
│
├── bridge/                  # IDE integration (VS Code, JetBrains)
├── coordinator/             # Multi-agent orchestration
├── plugins/                 # Plugin system
├── skills/                  # Skill system
├── server/                  # Server mode
├── remote/                  # Remote sessions
├── memdir/                  # Persistent memory directory
├── tasks/                   # Task management
├── state/                   # State management
│
├── voice/                   # Voice input
├── vim/                     # Vim mode
├── keybindings/             # Keybinding configuration
├── schemas/                 # Config schemas (Zod)
├── migrations/              # Config migrations
├── entrypoints/             # Initialization logic
├── query/                   # Query pipeline
├── ink/                     # Ink renderer wrapper
├── buddy/                   # Companion sprite (Easter egg 🐣)
├── native-ts/               # Native TypeScript utils
├── outputStyles/            # Output styling
└── upstreamproxy/           # Proxy configuration

Architecture

1. Tool System

src/tools/ — Every tool Claude can invoke is a self-contained module with its own input schema, permission model, and execution logic.

Tool	Description
File I/O
`FileReadTool`	Read files (images, PDFs, notebooks)
`FileWriteTool`	Create / overwrite files
`FileEditTool`	Partial modification (string replacement)
`NotebookEditTool`	Jupyter notebook editing
Search
`GlobTool`	File pattern matching
`GrepTool`	ripgrep-based content search
`WebSearchTool`	Web search
`WebFetchTool`	Fetch URL content
Execution
`BashTool`	Shell command execution
`SkillTool`	Skill execution
`MCPTool`	MCP server tool invocation
`LSPTool`	Language Server Protocol integration
Agents & Teams
`AgentTool`	Sub-agent spawning
`SendMessageTool`	Inter-agent messaging
`TeamCreateTool` / `TeamDeleteTool`	Team management
`TaskCreateTool` / `TaskUpdateTool`	Task management
Mode & State
`EnterPlanModeTool` / `ExitPlanModeTool`	Plan mode toggle
`EnterWorktreeTool` / `ExitWorktreeTool`	Git worktree isolation
`ToolSearchTool`	Deferred tool discovery
`SleepTool`	Proactive mode wait
`CronCreateTool`	Scheduled triggers
`RemoteTriggerTool`	Remote trigger
`SyntheticOutputTool`	Structured output generation

2. Command System

src/commands/ — User-facing slash commands invoked with / in the REPL.

Command	Description	Command	Description
`/commit`	Git commit	`/memory`	Persistent memory
`/review`	Code review	`/skills`	Skill management
`/compact`	Context compression	`/tasks`	Task management
`/mcp`	MCP server management	`/vim`	Vim mode toggle
`/config`	Settings	`/diff`	View changes
`/doctor`	Environment diagnostics	`/cost`	Check usage cost
`/login` / `/logout`	Auth	`/theme`	Change theme
`/context`	Context visualization	`/share`	Share session
`/pr_comments`	PR comments	`/resume`	Restore session
`/desktop`	Desktop handoff	`/mobile`	Mobile handoff

3. Service Layer

src/services/ — External integrations and core infrastructure.

Service	Description
`api/`	Anthropic API client, file API, bootstrap
`mcp/`	Model Context Protocol connection & management
`oauth/`	OAuth 2.0 authentication
`lsp/`	Language Server Protocol manager
`analytics/`	GrowthBook feature flags & analytics
`plugins/`	Plugin loader
`compact/`	Conversation context compression
`extractMemories/`	Automatic memory extraction
`teamMemorySync/`	Team memory synchronization
`tokenEstimation.ts`	Token count estimation
`policyLimits/`	Organization policy limits
`remoteManagedSettings/`	Remote managed settings

4. Bridge System

src/bridge/ — Bidirectional communication layer connecting IDE extensions (VS Code, JetBrains) with the CLI.

Key files: bridgeMain.ts (main loop) · bridgeMessaging.ts (protocol) · bridgePermissionCallbacks.ts (permission callbacks) · replBridge.ts (REPL session) · jwtUtils.ts (JWT auth) · sessionRunner.ts (session execution)

5. Permission System

src/hooks/toolPermission/ — Checks permissions on every tool invocation.

Prompts the user for approval/denial or auto-resolves based on the configured permission mode: default, plan, bypassPermissions, auto, etc.

6. Feature Flags

Dead code elimination at build time via Bun's bun:bundle:

import { feature } from 'bun:bundle'

const voiceCommand = feature('VOICE_MODE')
  ? require('./commands/voice/index.js').default
  : null

Notable flags: PROACTIVE · KAIROS · BRIDGE_MODE · DAEMON · VOICE_MODE · AGENT_TRIGGERS · MONITOR_TOOL

Key Files

File	Lines	Purpose
`QueryEngine.ts`	~46K	Core LLM API engine — streaming, tool loops, thinking mode, retries, token counting
`Tool.ts`	~29K	Base types/interfaces for all tools — input schemas, permissions, progress state
`commands.ts`	~25K	Command registration & execution with conditional per-environment imports
`main.tsx`	—	CLI parser + React/Ink renderer; parallelizes MDM, keychain, and GrowthBook on startup

Tech Stack

Category	Technology
Runtime	Bun
Language	TypeScript (strict)
Terminal UI	React + Ink
CLI Parsing	Commander.js (extra-typings)
Schema Validation	Zod v4
Code Search	ripgrep (via GrepTool)
Protocols	MCP SDK · LSP
API	Anthropic SDK
Telemetry	OpenTelemetry + gRPC
Feature Flags	GrowthBook
Auth	OAuth 2.0 · JWT · macOS Keychain

Design Patterns

Parallel Prefetch — Startup optimization

MDM settings, keychain reads, and API preconnect fire in parallel as side-effects before heavy module evaluation:

// main.tsx
startMdmRawRead()
startKeychainPrefetch()

Lazy Loading — Deferred heavy modules

OpenTelemetry (~400KB) and gRPC (~700KB) are loaded via dynamic import() only when needed.

Agent Swarms — Multi-agent orchestration

Sub-agents spawn via AgentTool, with coordinator/ handling orchestration. TeamCreateTool enables team-level parallel work.

Skill System — Reusable workflows

Defined in skills/ and executed through SkillTool. Users can add custom skills.

Plugin Architecture — Extensibility

Built-in and third-party plugins loaded through the plugins/ subsystem.

GitPretty Setup

Show per-file emoji commit messages in GitHub's file UI

# Apply emoji commits
bash ./gitpretty-apply.sh .

# Optional: install hooks for future commits
bash ./gitpretty-apply.sh . --hooks

# Push as usual
git push origin main

Contributing

Contributions to documentation, the MCP server, and exploration tooling are welcome. Changes to the archived snapshot under src/ are not the default contribution path. See CONTRIBUTING.md for guidelines.

Note: The src/ directory is the archived source snapshot and should generally remain unchanged.

Disclaimer

This repository archives a source snapshot reportedly exposed via Anthropic's npm distribution on 2026-03-31. It is provided for research, documentation, and exploratory tooling around the snapshot. The original Claude Code source remains the property of Anthropic, this is not an official release, and no rights to Anthropic's original code are granted by this repository. If you choose to use or redistribute any of the archived material, you are responsible for assessing the legal implications yourself. Contact nichxbt for any comments.

Name		Name	Last commit message	Last commit date
Latest commit History 2 Commits
.github		.github
.vscode		.vscode
docs		docs
mcp-server		mcp-server
scripts		scripts
src		src
web		web
.dockerignore		.dockerignore
.env.example		.env.example
.gitignore		.gitignore
.mcp.json		.mcp.json
CONTRIBUTING.md		CONTRIBUTING.md
Dockerfile		Dockerfile
LICENSE		LICENSE
README.md		README.md
Skill.md		Skill.md
agent.md		agent.md
biome.json		biome.json
bun.lock		bun.lock
bunfig.toml		bunfig.toml
gitpretty-apply.sh		gitpretty-apply.sh
package-lock.json		package-lock.json
package.json		package.json
server.json		server.json
tsconfig.json		tsconfig.json
vercel.json		vercel.json

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Repository files navigation

Claude Code Source Snapshot

Table of Contents

The Story: How the Source Code Leaked

Full Breakdown of Code Findings

1. The Terminal is a React App

2. Bun Runtime & Dead Code Elimination

3. Agent Swarms and Multi-Agent Architecture

4. Comprehensive Tool & Command Ecosystem

5. IDE "Bridge" System

What Is Claude Code?

Documentation

Explore with MCP Server

Install from npm

One-liner setup (from source)

Available tools & prompts

Custom source path / Remove

Directory Structure

Architecture

1. Tool System

2. Command System

3. Service Layer

4. Bridge System

5. Permission System

6. Feature Flags

Key Files

Tech Stack

Design Patterns

GitPretty Setup

Contributing

Disclaimer

About

Uh oh!

Contributors

Uh oh!

Languages

Folders and files

Latest commit

History

Repository files navigation

Claude Code Source Snapshot

Table of Contents

The Story: How the Source Code Leaked

Full Breakdown of Code Findings

1. The Terminal is a React App

2. Bun Runtime & Dead Code Elimination

3. Agent Swarms and Multi-Agent Architecture

4. Comprehensive Tool & Command Ecosystem

5. IDE "Bridge" System

What Is Claude Code?

Documentation

Explore with MCP Server

Install from npm

One-liner setup (from source)

Available tools & prompts

Custom source path / Remove

Directory Structure

Architecture

1. Tool System

2. Command System

3. Service Layer

4. Bridge System

5. Permission System

6. Feature Flags

Key Files

Tech Stack

Design Patterns

GitPretty Setup

Contributing

Disclaimer

About

Topics

Resources

License

Contributing

Uh oh!

Stars

Watchers

Forks

Contributors

Uh oh!

Languages