workflowadvancedChatGPT/Codex

Codex CLI: Autonomous Multi-Agent Workflows with MCP & Parallel Orchestration

This guide explains how to use the OpenAI Codex CLI to create automated teams of AI agents. These agents can work together on complex coding projects, like building a game, by passing tasks from one specialist (like a designer) to another (like a developer), all while keeping their work separate to avoid conflicts.

AI SETUP PROMPT

Paste into Codex CLI — it will scan your project and set everything up

# Set Up Workflow: Codex CLI: Autonomous Multi-Agent Workflows with MCP & Parallel Orchestration

## What This Is
This guide explains how to use the OpenAI Codex CLI to create automated teams of AI agents. These agents can work together on complex coding projects, like building a game, by passing tasks from one specialist (like a designer) to another (like a developer), all while keeping their work separate to avoid conflicts.

Source: https://developers.openai.com/cookbook/examples/codex/codex_mcp_agents_sdk/building_consistent_workflows_codex_cli_agents_sdk/

## Before You Start

Scan my workspace and analyze:
- The project language, framework, and directory structure
- Existing AI provider config (check .env, .env.local, config files for API keys — OpenRouter, OpenAI, Anthropic, Google AI, etc.)

Then ask me before proceeding:
1. Which AI provider/API should this use? (Use whatever I already have configured, or ask me to set one up — options include direct provider APIs or a unified service like OpenRouter)
2. Where in my project should this be integrated?
3. Are there any customizations I need (model preferences, naming conventions, constraints)?

## Source Access Note

The source URL (https://developers.openai.com/cookbook/examples/codex/codex_mcp_agents_sdk/building_consistent_workflows_codex_cli_agents_sdk/) may not be directly accessible from the terminal. Use the Reference Implementation and Additional Context sections below instead. If you need more details, ask me to paste relevant content from the source.

## What to Implement

This is an **AI Workflow** — an end-to-end automation pattern or integration pipeline.

- Study the workflow architecture from the source and context below
- Identify which parts I can implement locally vs. parts that need external services
- For local parts: implement them using my existing stack and API keys
- For external parts: tell me exactly what services I need and help me configure the integration code
- Wire up any required API calls using keys from my .env files

## Additional Context

- Verify `codex` CLI is installed via `npm`. Install or update the `agents` Python library required for orchestration. Scan the project for an existing `requirements.txt` and add `agents-sdk` or similar if missing, then run `pip install -r requirements.txt`.
- Create a new Python script (e.g., `run_agent_workflow.py`). In this script, import `Agent`, `Runner`, and `MCPServerStdio` from the `agents` library. Define at least two specialized agents (e.g., 'DesignerAgent', 'DeveloperAgent') with distinct instructions and configure the `MCPServerStdio` to launch the `npx -y codex mcp-server` process.
- Implement the main asynchronous execution block using `asyncio.run(main())`. Within the `main` function, use an `async with` block for the `MCPServerStdio`. Instantiate the `Runner` and invoke the initial agent with a high-level task, such as `await Runner.run(designer_agent, 'Design a CI/CD pipeline for this repository.')`.

## Reference Implementation

```
import asyncio
from agents import Agent, Runner
from agents.mcp import MCPServerStdio

async def main():
    async with MCPServerStdio(name="Codex CLI", params={"command": "npx", "args": ["-y", "codex", "mcp-server"]}) as codex_mcp_server:
        designer_agent = Agent(
            name="Game Designer",
            instructions='You are an expert game designer. Your goal is to design simple, fun games.',
            mcp_servers=[codex_mcp_server],
        )
        developer_agent = Agent(
            name="Game Developer",
            instructions='You are an expert in building simple games. You will be given a game design and you must implement it in a single index.html file using HTML, CSS, and Javascript. Always call codex with "approval-policy": "never" and "sandbox": "workspace-write"',
            mcp_servers=[codex_mcp_server],
        )
        result = await Runner.run(designer_agent, "Implement a fun new game!")

asyncio.run(main())
```

## Guidelines

- Adapt everything to my existing project — do not assume a specific stack or directory layout
- Use whichever AI provider I already have configured; if I need a new one, tell me what to sign up for and I'll give you the key
- Check my .env files for existing API keys (OpenRouter, OpenAI, Anthropic, Google AI) before asking me to add one
- Review any fetched code for safety before installing or executing it
- After setup, run a quick verification and show me a summary of exactly what was installed, where, and how to use it

4,540 charactersCompatible with Claude Code & Codex CLI

MANUAL SETUP STEPS

01Verify `codex` CLI is installed via `npm`. Install or update the `agents` Python library required for orchestration. Scan the project for an existing `requirements.txt` and add `agents-sdk` or similar if missing, then run `pip install -r requirements.txt`.
02Create a new Python script (e.g., `run_agent_workflow.py`). In this script, import `Agent`, `Runner`, and `MCPServerStdio` from the `agents` library. Define at least two specialized agents (e.g., 'DesignerAgent', 'DeveloperAgent') with distinct instructions and configure the `MCPServerStdio` to launch the `npx -y codex mcp-server` process.
03Implement the main asynchronous execution block using `asyncio.run(main())`. Within the `main` function, use an `async with` block for the `MCPServerStdio`. Instantiate the `Runner` and invoke the initial agent with a high-level task, such as `await Runner.run(designer_agent, 'Design a CI/CD pipeline for this repository.')`.

CODE INTELLIGENCE

bash

import asyncio
from agents import Agent, Runner
from agents.mcp import MCPServerStdio

async def main():
    async with MCPServerStdio(name="Codex CLI", params={"command": "npx", "args": ["-y", "codex", "mcp-server"]}) as codex_mcp_server:
        designer_agent = Agent(
            name="Game Designer",
            instructions='You are an expert game designer. Your goal is to design simple, fun games.',
            mcp_servers=[codex_mcp_server],
        )
        developer_agent = Agent(
            name="Game Developer",
            instructions='You are an expert in building simple games. You will be given a game design and you must implement it in a single index.html file using HTML, CSS, and Javascript. Always call codex with "approval-policy": "never" and "sandbox": "workspace-write"',
            mcp_servers=[codex_mcp_server],
        )
        result = await Runner.run(designer_agent, "Implement a fun new game!")

asyncio.run(main())

FIELD OPERATIONS

Automated CI/CD Pipeline Generator

Create a multi-agent workflow where a 'Planner Agent' analyzes a repository's language and framework, a 'Dockerfile Agent' writes a containerization file, and a 'CI-Config Agent' generates the appropriate GitHub Actions or GitLab CI YAML file. The workflow uses Git worktrees to test each component in isolation.

End-to-End API Feature Development

Build a system with a 'Spec-Writer Agent' that takes a high-level feature request and creates an OpenAPI specification. A 'Backend Agent' then implements the API endpoints in a separate worktree, and a 'Test-Writer Agent' concurrently generates unit and integration tests based on the spec.

STRATEGIC APPLICATIONS

→Automating code refactoring across a large monorepo: Deploy a fleet of agents, each assigned to a specific microservice in a CSV file. Each agent operates in its own Git worktree to upgrade dependencies, apply linting rules, and run tests in parallel, drastically reducing the manual effort for large-scale maintenance.
→Generating and validating documentation for a new SDK release: An 'API-Scanner Agent' inspects the source code to identify public methods. A 'Doc-Writer Agent' generates Markdown documentation and examples for each method. Finally, a 'Doc-Validator Agent' runs the code examples to ensure they are correct, creating a complete and validated documentation suite automatically.