What it is and who makes it
A terminal coding agent from OpenAI that can work against a local project and is also available through editor and app surfaces.
Maker or project: OpenAI. Classification: Terminal coding agent; this page treats the surrounding context, tools, permissions and verification as the relevant harness boundary.
Start the CLI inside a trusted project directory.
The agent reads relevant context and proposes or performs tool actions according to its approval mode.
It edits files or runs commands in the configured environment.
You inspect the diff and verification output before accepting the change.
Who it may fit
Pick an agent by its control surface and trust boundary, not only by its model label.
Useful scenarios
- Terminal-first software development
- Repository exploration, implementation and review tasks
- Developers who want an OpenAI coding-agent workflow close to Git and the shell
Trade-offs and checks
- Approval modes, instruction files and command flags change; use the current official CLI reference.
- Treat shell access and repository writes as privileged actions.
- Never allow an agent to deploy or rotate production credentials without an explicit, separate release process.
Interactive vs headless and CI
These modes change the approval, identity and observability boundary.
Interactive work is the natural terminal or app workflow: inspect the plan, approve actions and review the diff. For headless or CI use, follow the current documented non-interactive/automation surface and keep the workspace, permissions, time and test exit contract explicit.
Request-to-verification workflow
Use this mental model when debugging a surprising edit, tool call or cost.
Intent + context
Your prompt, repository instructions, selected files, history and environment become model context.
Tool + permission
The harness proposes a read, edit, shell, web, MCP or other action; policy decides whether it can run.
Result + verification
Command output, compiler errors, tests and your review return to the loop before acceptance.
Setup outline
Start with the smallest safe experiment, then follow the current official documentation for the real installation.
# Install path documented by the project npm install -g @openai/codex # Start in a repository cd your-project codex # Begin with a bounded task codex "Explain the request flow. Do not edit files."
Instructions, skills, MCP and hooks
These surfaces can shape context or add actions; they are part of the trust boundary.
Use the current Codex documentation for repository instruction files, skills, MCP and hooks. Treat instruction precedence and any extension that can run commands or read files as part of the permission model, not as harmless prompt text.
Official sources
Use these sources for current installation, capabilities, plans, releases and security notices.
Reviewed 2026-09-26. This date is a change-tracking note, not a promise that the product remains unchanged.
Related learning
Related agent guides
Want the foundations behind the workflow?
Study AI fundamentals, prompt design, AI-assisted development, agents and MCP without a forced linear course.