Provides AI-powered coding assistance through Codex CLI, enabling technical consultation, code review, and code explanation using OpenAI's models.
Click on "Deploy Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type @ followed by the MCP server name and your instructions, e.g., "@codex-cli-architect-mcp Review the auth middleware for security flaws"
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
An MCP (Model Context Protocol) server implementation for OpenAI's Codex CLI, enabling AI-powered coding assistance through standardized tool interfaces.
Technical Consultation: Get expert advice on architecture, design patterns, and technology choices
Code Review: Comprehensive code quality and security analysis
Code Explanation: Clear explanations of code snippets and programming concepts
Read-Only File Access: Can read files in the current workspace (directory and subdirectories)
Write Protection: Uses Codex CLI's read-only sandbox mode to prevent file modifications
Sandboxed Execution: Runs with platform-specific sandboxing (Apple Seatbelt on macOS, Landlock/seccomp on Linux)
No Code Modification: Focuses on analysis and advice, not code generation
Related MCP server: Codex MCP Server
Node.js 18+
Codex CLI installed and logged in
Register the MCP server with Claude Code:
# Using NPM package
claude mcp add codex -- npx codex-cli-architect-mcp
# With custom model configuration
claude mcp add codex -e CODEX_MODEL=gpt-5 -- npx codex-cli-architect-mcp
# With OpenAI API key
claude mcp add codex -e OPENAI_API_KEY=your-api-key-here -- npx codex-cli-architect-mcpConfigure the MCP server in your client's settings file:
{
"mcpServers": {
"codex": {
"command": "npx",
"args": ["codex-cli-architect-mcp"],
"env": {
"OPENAI_API_KEY": "your-api-key-here",
"CODEX_MODEL": "gpt-5",
"CODEX_TIMEOUT": "300000"
}
}
}
}The MCP server provides the following tools:
codex_consult - Technical consultation for architecture, design patterns, and implementation strategies
codex_review - Comprehensive code review for quality, security, and best practices
codex_explain - Get clear explanations of code snippets or programming concepts
These tools are exposed through the MCP protocol and can be used by any MCP-compatible client.
Environment variables:
OPENAI_API_KEY - OpenAI API key for authentication (optional, uses ChatGPT login if not provided)
CODEX_MODEL - Model to use (default: gpt-5)
CODEX_TIMEOUT - Execution timeout in milliseconds (default: 300000)
# Install dependencies
npm install
# Build
npm run build
# Development mode
npm run dev
# Run tests
npm testMIT
Technical consultation for architecture decisions, design patterns, technology choices, and implementation strategies
Inputs: topic (required), question (required), model, context, +1 more.
Explain code, features, or changes. Codex examines the implementation, Git history, or conceptual design to provide comprehensive explanations
Inputs: target (required), model, languages.
Review code for a feature, component, or Git changes. Codex analyzes the relevant implementation or modified files, providing contextual feedback and improvement suggestions
Inputs: target (required), focus, model, languages.
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
Scored across 3 tools
The three tools are largely distinct: consult is clearly separate, while explain and review both analyze code but differ in intent (understanding vs. critique). This boundary is clear enough for most cases, though explain and review could still be confused in edge scenarios.
All tool names follow the same predictable pattern: codex_ prefix plus a lowercase imperative verb in snake_case. The naming is consistent and each verb clearly reflects the tool's purpose.
Three tools is at the low end of the typical range, but each covers a distinct architect-focused activity: explain, consult, and review. The count is reasonable for a narrow assistant, though a broader architecture server could justify more tools.
The toolset covers core advisor-style workflows: understanding code, getting technical guidance, and reviewing changes. It lacks generative or planning tools, but that appears acceptable for an architect/consultant role rather than an active coding agent.
An MCP server that gives your AI access to the source code and docs of all public github repos
Driflyte MCP server which lets AI assistants query topic-specific knowledge from web and GitHub.
An MCP server that integrates with Discord to provide AI-powered features.
