OpenAI Codex
Coding agent and code generation assistant available via ChatGPT subscriptions and the OpenAI API with IDE CLI and web access for development tasks.
Overview
OpenAI Codex is a modern coding agent that plans runs and reviews changes across a project while keeping a verifiable log of actions. It writes functions explains diffs proposes fixes and can operate inside terminals IDEs and the web. Developers can tag the agent in pull requests for review generate tests or refactor modules with interactive guidance.
The latest release ships with GPT 5 Codex models that focus on software workflows and integrate with the Responses API. Access is delivered through ChatGPT paid tiers for individuals teams and enterprises and via API for programmatic use. The Codex CLI and extensions connect repositories to a cloud execution environment that runs tasks in parallel and returns artifacts and logs.
Codex supersedes the legacy Codex 2021 models that were retired and replaces them with a unified agentic experience across devices and accounts so teams can standardize coding help without juggling separate tools.
Key features
- Agentic coding sessions in terminal IDE and web with logs and artifacts
- GPT 5 Codex models focused on code review generation and refactoring
- Pull request reviews with inline suggestions and explainers
- Tests and bug fixes drafted from failing outputs and traces
- CLI and extensions to connect repos private or cloud sandboxes
- Responses API access to Codex models for programmatic control
- Works with ChatGPT accounts across Plus Pro Team and Enterprise
- Structured action history for audit friendly workflows
Best for
- Draft new features from structured tickets with commit level traceability
- Request refactors to modern patterns while preserving behavior
- Generate tests from examples and failing logs to raise coverage
- Review pull requests with inline reasoning and citation to changes
- Explain unfamiliar code paths during onboarding or audits
- Automate repetitive tasks like renames and boilerplate creation
- Spin up prototypes that call external APIs and verify outputs
- Use the API to script batch migrations with guardrails
Capabilities
Agentic sessions
Run tasks that plan change test and report while keeping artifacts and a clear action trail.
Pull requests
Tag the agent to summarize diffs point to risks and suggest focused changes.
Structure and tests
Upgrade patterns add tests and migrate modules with prompts tied to repository context.
API and CLI
Script repetitive edits and verifications using Codex in the Responses API and CLI.
Frequently Asked Questions
How do I get access and what is the starting price?
Codex access is included with paid ChatGPT subscriptions starting at ChatGPT Plus for $20 per month with higher tiers and API usage for advanced needs.
Was the old Codex API discontinued?
Yes, the legacy Codex models were retired and replaced by newer GPT based coding models and the Codex agent experience.
Can Codex run in my IDE or terminal?
Yes, official extensions and the Codex CLI support terminals IDEs and the web with repository aware context.
Is there an API for automation?
Yes, Codex models are available in the Responses API so teams can automate migrations reviews and checks.
How does Codex handle privacy and logs?
Codex produces action logs and artifacts and enterprise plans offer additional controls retention and privacy features
