Table of content
MCP servers for Codex integration
Installation
npx claude-plugins install @udecode/dotai/codex
Contents
Files: README.md
Documentation
MCP servers for Codex integration with Claude Code.
Overview
This plugin provides MCP (Model Context Protocol) servers that enable Claude Code to invoke Codex for complex tasks. This allows you to leverage different AI models within your Claude Code workflow.
Installation
/plugin install codex@dotai
Then restart Claude Code.
Features
GPT MCP Server
Provides access to GPT-5 with high reasoning effort for general-purpose tasks:
mcp__gpt__codex
Codex MCP Server
Provides access to GPT-5-Codex with high reasoning effort for coding tasks:
mcp__codex__codex
Prerequisites
You must have the codex CLI tool installed and configured:
# Install codex
npm install -g @openai/codex-cli
Usage
Once the plugin is installed and Claude Code is restarted, the MCP servers will be available as tools in Claude Code’s toolkit. Claude will automatically use these servers when appropriate for complex tasks.
Example Use Cases
- GPT Server: General problem-solving, analysis, research
- Codex Server: Code generation, debugging, refactoring
Configuration
The plugin uses the following default configuration:
- Model:
gpt-5(GPT) /gpt-5-codex(Codex) - Reasoning Effort:
high - Protocol: stdio (standard input/output)
These settings are optimized for quality output and can be modified in .mcp.json if needed.
Troubleshooting
MCP servers not appearing:
- Ensure codex CLI is installed:
which codex - Verify API key is configured:
codex config list - Restart Claude Code after installation
Permission errors:
- Make sure codex is executable:
chmod +x $(which codex) - Check that your OpenAI API key has proper permissions
License
MIT