OpenAI Codex: Build and Maintain Software with AI Coding Agents

For software engineers who already work with Git, tests and pull requests. Learn where OpenAI Codex runs and what it costs, then set up the Codex CLI with a sandbox, approval policy, AGENTS.md, configuration and MCP. Use Codex to plan and build features test by test, to refactor, upgrade and fix bugs, and to run cloud tasks and code review on pull requests. Review every agent change, keep secrets out of reach and gate Codex in CI. Then build on the OpenAI API: the Responses API, function calling and structured outputs, the Agents SDK with handoffs and guardrails, file search or your own retrieval, what it costs, and an IT service desk agent with a person approving every change. Finish with two hands on labs on small made up Python repositories: fix two bugs and build a feature, then extend a service desk triage agent, each ending in reviewed branches.