Codex for Developers.

You will learn how to run Codex as a professional tool across its surfaces, write AGENTS.md files that change behaviour, set sandbox and approval policy on purpose, extend it with MCP servers and skills, wire it into git and CI, and build a first agent loop on the Responses API and the Agents SDK. End state is an AGENTS.md committed to your most important repo, one real backlog item shipped through Codex with review, one MCP server connected, one CI job running codex exec, and a working agent loop with one tool of your own. After this module you can delegate scoped engineering work with confidence, structure the delegation so your team can reuse it, draw the autonomy line in a shared codebase, and move from using an agent to building one.

• OpenAI Codex documentation, overview, the surfaces and where each runs: https://learn.chatgpt.com/codex
• OpenAI Codex documentation, "AGENTS.md", discovery, merging and the size limit: https://learn.chatgpt.com/codex/agent-configuration/agents-md
• OpenAI Codex documentation, "Agent approvals and security", sandbox modes, approval policies, permission profiles and auto-review: https://learn.chatgpt.com/docs/agent-approvals-security
• OpenAI Codex documentation, "Codex GitHub Action", running codex exec in CI without exposing the key: https://learn.chatgpt.com/docs/github-action
• OpenAI API documentation, "Responses vs Chat Completions", why Responses is the recommended base for agents, and "Function calling", the loop shown above: https://developers.openai.com/api/docs/guides/responses-vs-chat-completions and https://developers.openai.com/api/docs/guides/function-calling
• OpenAI Agents SDK documentation, primitives, install and the hello-world example: https://openai.github.io/openai-agents-python/