MauricioMilano/coder-api

Coder-API

A powerful backend service that enables AI assistants to autonomously manage, modify, and execute code projects through secure sandboxed environments.

Coder-API bridges the gap between AI models and code execution, providing a comprehensive toolkit for project management, file operations, and command execution. Built with dual protocol support (REST API + Model Context Protocol), it seamlessly integrates with ChatGPT, GitHub Copilot, and other AI assistants to enable autonomous coding workflows.

Key Capabilities:

• 🔧 Project Management: Clone repositories, create projects, manage workspaces
• 📁 File Operations: Read, write, patch, and organize code files with precision
• ⚡ Code Execution: Run bash commands and scripts in isolated project environments
• 🤖 AI Design: Built with MCP support with automatic tool discovery for LLMs
• 🛡️ Secure Sandbox: All operations confined to designated workspace directories
• 🌐 Flexible Deployment: Local development, cloud platforms, or tunneled access

Features

The server provides the following capabilities:

Project Management

• create-project - Create new projects from Git repositories, archives, or empty directories
• rename-project - Rename existing projects
• list-projects - Get a list of all projects
• get-project-details - Get detailed information about a specific project

File Operations

• get-file - Read file contents with support for text and base64 encoding
• create-file - Create new files with overwrite protection
• delete-file - Delete files or directories with recursive options
• patch-file - Apply modifications using various patch operations (diff, replace, lines, insert, code_block)

System Operations

• run-bash - Execute bash commands in project directories with timeout and environment controls
• list-filetree - Browse project file structures with configurable depth and entry limits

How we build it

• Express server with dual protocol support:
  • REST API for traditional HTTP-based project and file management
  • MCP (Model Context Protocol) for seamless LLM integration
• OpenAPI contract (openapi.json)
• Multiple transport options for MCP (HTTP and Server-Sent Events)

Getting Started

For detailed local setup instructions, see .

Quick start:

• Node.js >= 20
• pnpm (recommended)
• Clone, install dependencies, configure .env, and run pnpm dev

Deployment

For detailed setup and deployment instructions, see our deployment guides:

• - Local development setup and testing
• - Deploy on your own VPS or server
• - Deploy on Render, Heroku, Easypanel, and other cloud services
• - Use ngrok or other tunneling services for development

To allow external services (like ChatGPT or other AI assistants) to access your Coder-API, you need to make it accessible over the internet using one of the deployment methods above.

---

API Documentation

Coder-API provides two main interfaces:

REST API

Traditional HTTP-based API for project and file management.

• - Complete REST API guide with examples
• OpenAPI Contract: /openapi endpoint or openapi.json file

Model Context Protocol (MCP)

Standardized protocol for AI assistant integration.

• - MCP protocol documentation and examples

How To Use

• - Using MCP with ChatGPT
• - Using REST API with ChatGPT
• - Using with VS Code

Security Notes

• All file operations are confined to WORKSPACE_ROOT/{projectId}.
• Bash commands are not fully network-isolated (MVP). Use in a controlled environment.
• See deployment documentation for security considerations specific to each deployment method.

Contributing

1. Fork and clone the repo.

2. Create a feature branch.

3. Open a pull request with a clear description.