⏳ In Brief
- Codex is an AI coding agent included with paid ChatGPT plans.
- Works in terminal, IDE, web, and mobile, plus delegated cloud tasks.
- Adds code review automation and GitHub task handoffs with tagging.
- Access via ChatGPT sign-in, with usage limits that vary by plan.
- Recommends GPT-5 as default, with adjustable reasoning levels.
OpenAI ships Codex as an everywhere coding agent for ChatGPT
OpenAI introduced Codex as a single agent that pairs in your terminal or IDE, then continues work in the cloud when delegated. It is included with ChatGPT Plus, Pro, Team, Edu, and Enterprise plans.
The agent edits files, runs commands, and executes tests from prompts or specs. Users can keep flow locally, then offload longer jobs to Codex cloud, reviewing diffs before merging changes to the main branch.
Codex is positioned as “one agent for everywhere you code,” available through paid ChatGPT plans, with install via the Codex CLI and an IDE extension for VS Code, Cursor, and Windsurf.
We’re releasing new Codex features to make it a more effective coding collaborator:
– A new IDE extension
– Easily move tasks between the cloud and your local environment
– Code reviews in GitHub
– Revamped Codex CLIPowered by GPT-5 and available through your ChatGPT plan.
— OpenAI Developers (@OpenAIDevs) August 27, 2025
What Codex actually does across tools and workflows
Codex navigates a repo, applies multi-file edits, and writes unit tests that run in an isolated sandbox. You can pull results to your machine, or leave work in the cloud for asynchronous review.
It also accelerates code reviews. Teams can enable automatic comments on pull requests, then tag @codex to trigger tasks from GitHub, including adding tests and addressing requested changes.
Where Codex runs today
- Terminal and IDE pairing for local tasks
- Delegated cloud jobs with sandboxed execution
“Codex runs in the background so you can stay in flow and get more done.”
Access, setup, and usage limits by plan
Subscribers sign in to Codex with their ChatGPT account through the CLI, IDE extension, or web client. Team and Enterprise admins can enable workspace access and fine-tune controls centrally.
Plan limits differ. Plus and Team users get local tasks with a rolling window, and cloud tasks with generous allowances. Pro raises local headroom substantially, while Enterprise and Edu can draw from shared credits.
If you previously used an API key, update the CLI, run codex logout, then authenticate with ChatGPT to switch to subscription-based access and current limits.
Security, compliance, and data handling
Enterprise features include RBAC, workspace-level enablement, and logging through a Compliance API for web and cloud usage. Local sessions run on your own environment, outside those central logs.
Data use varies by plan. Business tiers default to no training on inputs or outputs, while Plus and Pro can disable training in ChatGPT data controls. Residency and retention policies apply across Codex services.
Models, compatibility, and where to start
The CLI and IDE extension default to GPT-5, with selectable reasoning levels to match task complexity. You can also authenticate with an API key to use older models when needed.
Codex supports VS Code, popular forks, and works from your terminal in other IDEs. Mobile access lets you start tasks, check reviews, and merge on the go through the ChatGPT app.
“With your ChatGPT account, you can move work between local and Codex’s cloud sandbox without losing state.”
What to watch next for teams adopting Codex
Teams should trial automated reviews on a subset of repositories, measure defect rates, and track cycle-time for PRs that include Codex feedback. Early guardrails will reveal where human oversight adds the most value.
Administrators can stage RBAC, confirm Compliance API coverage for cloud runs, and document data policies for contributors. Clear onboarding reduces friction as Codex moves from evaluation to daily use.
Conclusion
Codex reframes AI coding as a persistent agent, present in local tools and the cloud, tied to ChatGPT authentication. That reach enables practical edits, tests, and reviews without constant context resets.
Its inclusion across paid ChatGPT tiers lowers the barrier to team-wide pilots. With plan-specific limits, enterprise controls, and GPT-5 defaults, the question now is workflow change, not access to core capability.
📈 Trending News
28th August 2025
- WhatsApp’s AI “Writing Help” rolls out to fix tone, typos, and phrasing
- ‘What if’ AI video of Mount Fuji erupting over Tokyo
- Meet Plaud’s Note Pro — a card-sized AI recorder with a display
- Some 911 centers are letting AI answer the emergency calls
- Anthropic says Agentic AI is now weaponized
For more AI stories, visit AI News on our site.
