MCP

This project successfully reconstructs the complete source code of the CLI tool by restoring the source maps left behind during the Claude Code 2.1.88 release. Developers can use this to delve into its terminal UI built on React and Ink, its complex command system, and its MCP integration implementation. The repository aims to provide a technical architectural reference to help users understand the internal logic and engineering design of Claude Code.

Context Mode is an MCP server designed to prevent context window exhaustion by offloading raw data into a sandboxed SQLite database. It tracks session events and uses BM25 search to retrieve only relevant information, ensuring the LLM maintains continuity during conversation compaction. Additionally, it encourages a code-first approach where agents write scripts to process data, significantly reducing token consumption.

Claude Context is an MCP plugin that enables semantic code search for AI coding agents by indexing your entire codebase into a vector database. It significantly reduces costs and improves retrieval quality by providing only relevant code snippets to the AI instead of loading entire directories. The tool supports incremental indexing, AST-based code chunking, and integrates seamlessly with various AI assistants and IDEs.

Toprank is a Claude Code plugin that provides AI agents with direct access to Google Search Console and Google Ads for data-driven optimization. It enables users to perform automated audits, identify wasted ad spend, and implement technical SEO improvements through simple CLI commands. The tool also supports cross-model reviews via Gemini and integrates with various CMS platforms to streamline content and performance management.

GitNexus indexes codebases into a comprehensive knowledge graph to provide AI agents with deep architectural context. It offers both a CLI with MCP support for local development and a web-based UI for visual exploration and chat. By mapping dependencies and execution flows, it ensures AI agents maintain high accuracy during complex coding tasks.

code-review-graph builds a structural map of your codebase using Tree-sitter to provide AI assistants with precise, context-aware information. By tracking changes incrementally and calculating the blast radius of modifications, it significantly reduces token consumption during code reviews. The tool integrates seamlessly with various AI coding platforms via the Model Context Protocol to ensure only relevant code is analyzed.

Chrome DevTools for Agents is an MCP server that enables AI coding assistants to control and inspect live Chrome browser instances. It provides a comprehensive suite of tools for browser automation, performance analysis, and in-depth debugging. The project supports seamless integration with various AI coding platforms to enhance developer workflows through reliable browser interaction.

This project successfully restored approximately 700,000 lines of original code by parsing the source map files left in the Claude Code 2.1.88 npm package. It aims to provide developers with a reference platform for in-depth research into the Claude Code CLI architecture, command system, and MCP implementation. All restored content is for technical archiving and source code reading research purposes only, does not represent an official stance, and users should be aware of relevant legal risks.

The @ant-design/cli is a fully offline command-line tool that provides developers and code agents with instant access to Ant Design component knowledge. It supports multiple versions of antd, offering features like API lookups, project diagnostics, and automated migration guidance. The tool is optimized for agent integration through structured JSON output and an MCP server for seamless IDE connectivity.

This project contains the complete leaked source code for Anthropic's official Claude Code CLI tool, which was discovered on March 31, 2026, via map files in the npm package. The repository provides the original source code built with TypeScript and Bun, accompanied by detailed architectural documentation and exploration guides. Users can interactively delve into the tool's internal implementation and design patterns through the built-in MCP server.