OpenAI Codex MCP Server

@Tomatio13

About OpenAI Codex MCP Server

An MCP server to wrap the OpenAI Codex CLI for use with Claude Code

#openai-codex #mcp-server #ai-integration

Config

Add this server to your MCP-compatible client using the configuration below.

{
  "mcpServers": {
    "openai-codex-mcp-tomatio13": {
      "command": "python",
      "args": [
        "codex_server.py",
        "--mode",
        "stdio"
      ]
    }
  }
}

Tools

No tools detected

We auto-extract tools from the README. The maintainer can list them under a ## Tools heading to populate this section.

Overview

What is OpenAI Codex MCP Server?

OpenAI Codex MCP Server wraps OpenAI Codex CLI as an MCP (Model Context Protocol) server for integration with Claude. It provides tools to generate, explain, debug, refactor, and secure code using Codex capabilities directly from Claude.

How to use OpenAI Codex MCP Server?

Clone the repository, run ./setup.sh, then start the server using ./run.sh (stdio mode for Claude Desktop) or ./run_sse.sh [port] [host] (SSE mode for Web API). Configure Claude Desktop to use the server by adding an entry with command python /path/to/codex_server.py --mode stdio to the Claude Desktop configuration file. After setup, invoke tools like @openai-codex codex_agent or @openai-codex codex_interactive with relevant parameters.

Key features of OpenAI Codex MCP Server

Provides codex_agent and codex_interactive tools
Supports code generation, explanation, debugging, refactoring
Includes security analysis, test creation, documentation generation
Multimodal support (image input)
Multiple AI provider support (OpenAI, Azure, Gemini, Ollama, etc.)
3‑level automation modes: suggest, auto‑edit, full‑auto
Runs in both stdio and SSE modes

Use cases of OpenAI Codex MCP Server

Generate new code from natural language prompts
Debug and fix bugs in existing codebases
Perform security analysis and apply fixes
Write or improve documentation and tests
Explain complex code sections interactively

FAQ from OpenAI Codex MCP Server

What are the prerequisites for using this server?

You need Node.js with npm install -g @openai/codex, Python 3.12 or higher, and a .env file copied from .env.example.

What modes does the server support?

The server supports stdio mode (for Claude Desktop) and SSE mode (for Web API integration).

What tools does the server provide?

Two tools: codex_agent (comprehensive Codex features) and codex_interactive (interactive sessions).

What approval modes are available?

Three modes: suggest (only suggests changes, requires approval), auto-edit (automatic file read/write, asks for shell commands), and full-auto (full autonomy with network‑disabled sandbox).

How does security work in full‑auto mode?

The server runs full-auto mode in a network‑disabled sandbox. File operations are limited to the current working directory, and all shell commands require explicit approval unless in full-auto mode.

Frequently asked questions

What are the prerequisites for using this server?

You need Node.js with `npm install -g @openai/codex`, Python 3.12 or higher, and a `.env` file copied from `.env.example`.

What modes does the server support?

The server supports stdio mode (for Claude Desktop) and SSE mode (for Web API integration).

What tools does the server provide?

Two tools: `codex_agent` (comprehensive Codex features) and `codex_interactive` (interactive sessions).

What approval modes are available?

Three modes: `suggest` (only suggests changes, requires approval), `auto-edit` (automatic file read/write, asks for shell commands), and `full-auto` (full autonomy with network‑disabled sandbox).

How does security work in full‑auto mode?

The server runs `full-auto` mode in a network‑disabled sandbox. File operations are limited to the current working directory, and all shell commands require explicit approval unless in `full-auto` mode.

Comments

Basic information

More AI & Agents MCP servers

1Panel

1Panel-dev

🔥 1Panel is a modern, open-source VPS control panel — and the only one with native AI agent support. Run Ollama models, deploy OpenClaw agents, and manage your entire server stack from one clean web interface.

36.1K