Agentic Developer

Provides an MCP endpoint that wraps Codex CLI and exposes it through the TeaBranch middleware for MCP-based workflows.

python

42

GitHub Stars

python

Language

2 months ago

First Indexed

3 weeks ago

Catalog Refreshed

Documentation & install

Readme and setup notes from the catalogue, plus a client-ready config you can copy for your MCP host.

Installation

Add the following to your MCP client configuration file.

Configuration

View docs

{
  "mcpServers": {
    "teabranch-agentic-developer-mcp": {
      "command": "python",
      "args": [
        "-m",
        "mcp_server"
      ]
    }
  }
}

You deploy a specialized MCP server that exposes Codex CLI functionality through a lightweight Express-based wrapper and an open-responses middleware, so you can send model-context operations to Codex via MCP clients with minimal setup. This setup is useful when you want programmatic code analysis and generation workflows bridged through a consistent MCP interface.

How to use

Start one of the MCP servers and point your MCP client at it. You can run the server locally to accept MCP requests, then send prompts to Codex through the standard MCP transport you choose (stdio or SSE). The server acts as a bridge: your client talks MCP, the bridge forwards requests to Codex, and responses come back formatted for MCP.

How to install

Prerequisites you need to satisfy before starting are: node version 22.x for Codex integration and Python for the standard MCP server runtime. You will also need npm and pip for installing dependencies and running the server in different modes.

# Docker (Recommended) 
# Clone this project
git clone https://github.com/yourusername/codex-mcp-wrapper.git
cd codex-mcp-wrapper

# Start the services
./start.sh

# This will start:
# Codex MCP wrapper on port 8080
# open-responses-server on port 3000

Manual installation steps let you run the server directly without Docker. Follow these commands to install dependencies and start the MCP bridge yourself.

# Install dependencies
npm install

# Install Codex CLI globally
npm install -g @openai/codex

# Start the MCP server
node mcp-server.js
# Install the package in development mode
pip install -e .

Additional configuration and notes

The project includes a configuration path for MCP tools under a dedicated directory. You can register available MCP tools by placing a configuration file named mcps.json under the .agent/ directory. Codex will load this configuration automatically when the server starts.

{
  "mcpServers": {
    "agentic-developer-mcp": {
      "url": "..."
    }
  }
}

If you prefer starting the server in Python, you can run the standard MCP server directly, which is compatible with the MCP Python SDK.

Examples of usage patterns

You can use either stdio or SSE transport when launching the server to communicate with MCP clients. Choose the transport that matches your client capabilities and network environment.

Available tools

run_codex

Clones a repository, checks out a branch, navigates to a folder, and runs Codex with a given request.

clone_and_write_prompt

Clones a repository, reads the system prompt file, parses model information, writes a prompt file, and invokes Codex with the extracted model