This project provides an MCP server implementing Claude Code-like capabilities, enabling LLMs like Claude to directly execute instructions for modifying and improving project files. It targets developers and researchers seeking to integrate LLM-powered code manipulation into their workflows, offering benefits such as automated code refactoring, bug fixing, and feature implementation.

How It Works

The implementation leverages the Model Context Protocol (MCP) to facilitate seamless integration with MCP clients. It acts as an intermediary, translating LLM instructions into executable actions on the local filesystem and codebase. Key features include code understanding via file analysis and pattern searching, targeted code modification with permission handling, enhanced command execution, and robust file operations with security controls. It supports multiple LLM providers via LiteLLM and offers specialized tools like grep_ast for context-aware code searching and notebook_edit for Jupyter notebook manipulation.

Quick Start & Requirements

Installation and configuration details are available in INSTALL.md. A tutorial for version 0.3 can be found in TUTORIAL.md.

Highlighted Details

• Implements a comprehensive suite of tools for code interaction, including read, write, edit, multi_edit, directory_tree, grep, content_replace, grep_ast, run_command, notebook_read, notebook_edit, think, dispatch_agent, batch, todo_write, and todo_read.
• Utilizes ripgrep for high-performance pattern searching (grep).
• Supports AST context for code searching (grep_ast) and Jupyter notebook cell manipulation.
• Includes agent delegation for concurrent task execution and support for multiple LLM providers via LiteLLM.

Maintenance & Community

The project is open-source, with contribution guidelines provided for forking, branching, committing, and opening pull requests.

Licensing & Compatibility

This project is licensed under the MIT License, permitting commercial use and integration with closed-source projects.

Limitations & Caveats

The implementation includes permission prompts for file modifications and command execution, restricts access to specified directories, and performs input validation and sanitization to enhance security.