MCP.so
登录

Codex Mcp Server

@x51xxx

关于 Codex Mcp Server

暂无概览

基本信息

分类

开发工具

许可证

MIT

运行时

node

传输方式

stdio

发布者

x51xxx

提交者

Тарас Тріщук

配置

使用下面的配置,将此服务器添加到你的 MCP 客户端。

{
  "mcpServers": {
    "codex-cli": {
      "command": "npx",
      "args": [
        "-y",
        "@trishchuk/codex-mcp-tool"
      ]
    }
  }
}

工具

4

Execute Codex CLI with file analysis, models, sessions

Generate ideas with SCAMPER, design-thinking, etc.

View/delete/clear conversation sessions

Diagnose CLI installation, version, features

概览

What is Codex MCP Server?

An MCP server that connects Claude or Cursor to Codex CLI, enabling code analysis via @ file references, multi-turn conversations, sandboxed edits, and structured change mode. It is for developers who want to leverage Codex CLI’s capabilities through an MCP-compatible client.

How to use Codex MCP Server?

Install and add via claude mcp add codex-cli -- npx -y @trishchuk/codex-mcp-tool or configure in claude_desktop_config.json. Prerequisites: Node.js 18+ and a working Codex CLI installation (authenticated). Invoke tools like ask-codex, brainstorm, list-sessions, or health.

Key features of Codex MCP Server

  • File analysis with @ file references
  • Multi-turn conversations with workspace isolation
  • Native resume using codex resume (CLI v0.36.0+)
  • Local OSS model support via Ollama or LM Studio
  • Sandboxed code execution with --full-auto
  • Structured brainstorming with SCAMPER framework

Use cases of Codex MCP Server

  • Analyze code architecture by referencing files via @src/
  • Conduct multi-turn code review sessions using session IDs
  • Run untrusted code safely in sandbox mode
  • Generate refactoring ideas with structured brainstorming
  • Use local models for code analysis without sending data to OpenAI

FAQ from Codex MCP Server

What are the runtime dependencies?

Node.js 18+ and a locally installed and authenticated Codex CLI are required.

How do I configure multi-turn conversations?

Pass a sessionId parameter in your prompt. Sessions persist for 24 hours by default (configurable via CODEX_SESSION_TTL_MS), with a maximum of 50 sessions.

What models are supported?

By default the CLI’s configured model is used. You can override with model names like gpt-5.5, gpt-5.4, gpt-5.4-mini, or use local OSS models via localProvider: "ollama" or "lmstudio".

Where does session data live?

Session data is stored locally on your machine under the Codex CLI workspace; no data is sent to external servers except through your chosen model provider.

What CLI versions are compatible?

Features work best with Codex CLI v0.36.0+ (sessions), v0.52.0+ (search), and v0.59.0+ (add-dirs, token limits).

评论

开发工具 分类下的更多 MCP 服务器