UNPKG

codex-mcp-server

Version:
238 lines (177 loc) 8.49 kB
# Codex MCP Server [![npm version](https://img.shields.io/npm/v/codex-mcp-server.svg)](https://www.npmjs.com/package/codex-mcp-server) [![npm downloads](https://img.shields.io/npm/dm/codex-mcp-server.svg)](https://www.npmjs.com/package/codex-mcp-server) [![license](https://img.shields.io/npm/l/codex-mcp-server.svg)](https://www.npmjs.com/package/codex-mcp-server) Advanced MCP server wrapper for OpenAI Codex CLI v0.50.0+ that provides enterprise-grade conversational AI coding assistance with session management, intelligent model selection, and native resume functionality. ```mermaid graph LR A[Claude Code] --> B[Codex MCP Server] B --> C[codex tool] B --> D[listSessions tool] B --> E[ping tool] B --> F[help tool] C --> G[Codex CLI v0.50.0+] C --> H[Session Storage] style A fill:#FF6B35 style B fill:#4A90E2 style C fill:#00D4AA style D fill:#00D4AA style E fill:#00D4AA style F fill:#00D4AA style G fill:#FFA500 style H fill:#9B59B6 ``` ## Prerequisites - **OpenAI Codex CLI v0.50.0+** must be pre-installed and configured - Install: `npm i -g @openai/codex` or `brew install codex` - **Setup**: Run `codex login --api-key "your-openai-api-key"` - ⚠️ **Breaking Change**: `OPENAI_API_KEY` environment variable is no longer supported - ⚠️ **Version Requirement**: v0.50.0+ required (see [Version Compatibility](#version-compatibility) below) - **Claude Code** installed ## Version Compatibility This MCP server requires **codex CLI v0.50.0 or later** due to the following changes: - **v0.50.0+**: Introduced `--skip-git-repo-check` flag (now required) - **v0.50.0+**: Removed `--reasoning-effort` flag (no longer supported) **If you have an older version of codex CLI**, you will need to upgrade: ```bash npm update -g @openai/codex ``` For detailed version compatibility information, see [docs/codex-cli-integration.md](docs/codex-cli-integration.md). ## Installation ### One-Click Installation #### VS Code [![Install in VS Code](https://img.shields.io/badge/VS_Code-Install_Codex_MCP_Server-0098FF?style=flat-square&logo=visualstudiocode&logoColor=white)](https://vscode.dev/redirect/mcp/install?name=codex-cli&config=%7B%22type%22%3A%22stdio%22%2C%22command%22%3A%22npx%22%2C%22args%22%3A%5B%22-y%22%2C%22codex-mcp-server%22%5D%7D) #### VS Code Insiders [![Install in VS Code Insiders](https://img.shields.io/badge/VS_Code_Insiders-Install_Codex_MCP_Server-24bfa5?style=flat-square&logo=visualstudiocode&logoColor=white)](https://insiders.vscode.dev/redirect/mcp/install?name=codex-cli&config=%7B%22type%22%3A%22stdio%22%2C%22command%22%3A%22npx%22%2C%22args%22%3A%5B%22-y%22%2C%22codex-mcp-server%22%5D%7D) #### Cursor [![Install in Cursor](https://img.shields.io/badge/Cursor-Install_Codex_MCP_Server-00D8FF?style=flat-square&logo=cursor&logoColor=white)](https://cursor.com/en/install-mcp?name=codex&config=eyJ0eXBlIjoic3RkaW8iLCJjb21tYW5kIjoibnB4IC15IGNvZGV4LW1jcC1zZXJ2ZXIiLCJlbnYiOnt9fQ%3D%3D) ### Manual Installation #### Claude Code ```bash claude mcp add codex-cli -- npx -y codex-mcp-server ``` #### Claude Desktop Add to your Claude Desktop configuration file: **macOS:** `~/Library/Application Support/Claude/claude_desktop_config.json` **Windows:** `%APPDATA%/Claude/claude_desktop_config.json` ```json { "mcpServers": { "codex-cli": { "command": "npx", "args": ["-y", "codex-mcp-server"] } } } ``` ## Usage in Claude Code Once installed, Claude Code can use these tools: ### `codex` - AI Coding Assistant Ask Codex to analyze code, generate solutions, or provide coding assistance with optional session support for conversational context. **Basic Usage:** ``` Use the codex tool to explain this function: [paste your code here] ``` **Advanced Usage:** ``` # Model selection (defaults to gpt-5-codex) Use codex with model "gpt-4" to analyze this complex algorithm # Reasoning effort control Use codex with reasoningEffort "high" for thorough code review # Session with model preferences (using advanced GPT-5 Codex by default) Use codex with sessionId "my-session" and model "gpt-4" to refactor this code # Continue conversation (uses native codex resume) Use codex with sessionId "my-session" to make it more efficient # Reset session context Use codex with sessionId "my-session" and resetSession true to start fresh analysis ``` **Parameters:** - `prompt` (required): Your coding question or request - `sessionId` (optional): Session ID for conversational context - `resetSession` (optional): Reset session history before processing - `model` (optional): Specify model to use (defaults to 'gpt-5-codex'). Options: 'gpt-5-codex', 'gpt-4', 'gpt-3.5-turbo' - `reasoningEffort` (optional): Control reasoning depth ('low', 'medium', 'high') ### `listSessions` - Session Management List all active conversation sessions with metadata including creation time, last access, and turn count. **Usage:** ``` Use listSessions to see all active coding sessions ``` ### `ping` - Connection Test Test if the MCP server is working properly. ### `help` - Codex CLI Help Get information about Codex CLI capabilities and commands. ## Example Workflows **Code Analysis:** ``` Please use the codex tool to review this TypeScript function and suggest improvements ``` **Conversational Code Development:** ``` # Start a session for complex refactoring Use codex with sessionId "refactor-auth" to analyze this authentication system # Continue building on the analysis Use codex with sessionId "refactor-auth" to implement the security improvements you suggested # Check session history Use listSessions to see all active development sessions ``` **Bug Fixing:** ``` Use codex to help debug this error: [error message] ``` **Code Generation:** ``` Ask codex to create a React component that handles file uploads ``` ## Advanced Features (Codex CLI v0.50.0+) ### Session Management **Key Features:** - **Native Resume**: Leverages `codex resume` for optimal conversation continuity - **Intelligent Defaults**: GPT-5-Codex model for superior coding assistance - **Fallback Context**: Manual context building when resume unavailable - **Session Persistence**: Context maintained across interactions (24hr TTL) - **Production-Ready**: Comprehensive error handling and graceful degradation - **Enterprise-Scale**: 54 tests covering all functionality and edge cases ### Model Selection & Control **Key Features:** - **Default GPT-5 Codex**: Automatically uses the latest `gpt-5-codex` model for optimal coding assistance - **Dynamic Models**: Choose between different models per request - **Reasoning Effort**: Control AI processing depth (low/medium/high) - **Per-Model Optimization**: Optimized prompts for different model families - **Flexible Configuration**: Mix and match model and session parameters ### Enhanced Authentication **Requirements:** - ⚠️ **Breaking Change**: Environment variable `OPENAI_API_KEY` no longer supported - **New Setup**: Run `codex login --api-key "your-key"` for authentication - **Credential Storage**: Credentials stored securely in `CODEX_HOME/auth.json` **Best Practices:** - Use sessions with native resume for complex, multi-step development tasks - **Default Model**: Uses `gpt-5-codex` by default for optimal coding assistance - Choose appropriate reasoning effort: `low` for quick answers, `high` for complex analysis - Select models based on task complexity: `gpt-5-codex` for coding (default), `gpt-4` for advanced reasoning, `gpt-3.5-turbo` for speed - Reset sessions when switching to unrelated topics - Check `listSessions` to manage active conversations ## Development ```bash # Install dependencies npm install # Development mode npm run dev # Build npm run build # Start built server npm start ``` ## Documentation ### 📚 Comprehensive Guides - **[Session Management](docs/session-management.md)** - Advanced session features and implementation details - **[Codex CLI Integration](docs/codex-cli-integration.md)** - v0.50.0+ features, breaking changes, and migration guide - **[API Reference](docs/api-reference.md)** - Complete tool documentation and usage examples ### 🔧 Development Resources - **Testing Suite**: 54 comprehensive tests covering all functionality - **TypeScript Support**: Full type definitions and IntelliSense - **Error Handling**: Robust error recovery and graceful degradation - **Performance**: Optimized for enterprise-scale usage ## License ISC