OpenAI's Codex Now Codes Autonomously While Your Mac Sits Locked
Goal Mode exits beta, Appshots brings one-keystroke screen context, and Codex can now keep coding after your Mac locks
Takeaways
- Goal Mode is GA: Goal mode exits beta and is now available across the Codex app, IDE extension, and CLI for multi-hour autonomous tasks.
- Appshots on macOS: Press both Command keys to instantly send any app window (screenshot + text) to a Codex thread, no copy-pasting required.
- Locked Mac support: Codex can now continue Computer Use tasks after your Mac screen locks, including triggered remotely from your phone.
- Shared plugins for teams: Business users can now distribute custom plugins across their workspace; enterprise early access available on request.
- Richer enterprise analytics: New breakdowns for active users, credits, tokens, runs, leaderboards, lines of code, and plugin usage, plus Analytics API updates.
- Advanced browser annotations: Directly adjust page elements in the in-app browser, preview changes instantly, and batch comments for Codex.
OpenAI's "Codex Thursday" dropped five updates at once, and the headline is a shift in how you think about AI-assisted coding. Goal mode is no longer an experimental feature and is available in the Codex app, IDE extension, and CLI. Paired with a new screen-capture shortcut called Appshots and the ability to keep working after your Mac locks, this release is less about making Codex smarter and more about making it autonomous enough to work while you're not watching.
From prompts to objectives
The old model for using Codex was transactional: send a message, wait for a result, send another. Goal mode changes that. Instead of sending Codex one instruction and waiting, you assign an objective and let it run across session breaks, token budget resets, and interruptions. With Goal mode, you can have Codex drive toward a specific objective for hours or even days.
The key mechanic is that the goal text acts as both the starting prompt and the completion criteria. Codex uses it to decide what to do next and whether the task is complete. You start Goal mode with /goal in the Codex app, IDE extension, or CLI. In the app, progress appears above the composer with controls to pause, resume, edit, or clear the goal.
Writing a good goal is the skill to develop here. Vague objectives produce vague results. The official docs suggest goals that include a specific outcome, measurable target, or test criteria. For example:
/goal Migrate this codebase from JavaScript to TypeScript.
# The app should compile in strict mode without explicit `any` type definitions.
/goal Reduce the time to interactive of the home page to below 1 second.You can also steer Codex mid-flight. Send follow-up messages to adjust constraints, such as asking Codex to use a particular library or avoid a specific approach. Use side chats when you want a status recap or explanation without interrupting the main task.
Appshots: context without the paste
The second big feature solves a friction point that's easy to overlook until you've felt it a dozen times. Every time you need to show Codex what's on your screen, you copy text, take a screenshot, describe the UI state. Appshots are now available in the Codex app on macOS. Press both Command keys to send the frontmost app window to Codex with a screenshot and available text, so Codex can work from context in another app without you copying, pasting, or describing it manually.
What makes this more than a screenshot tool is the text extraction. An appshot can include an image of the visible window, plus available text from that window, including visible text and text the app makes available outside the visible scroll area. So if you're looking at a long API reference page, Codex gets the full document, not just the visible portion.
Practical cases where this shines:
- An error panel throwing a stack trace you'd normally copy-paste
- An API reference open in a browser tab you want Codex to implement against
- A design mock in Figma you need Codex to match in code
- A settings panel or app state that's easier to show than describe
One limitation worth knowing: for some apps and websites, including Google Docs, Gmail, Google Sheets, and Google Slides, Codex may receive only the visible screenshot and may not receive the full document or off-screen text. For those, you'll want the matching plugin installed. Appshots are also macOS-only for now, with no Windows or Linux support yet.
Locked Mac, still working
The third feature is the one that raises eyebrows. Remote computer use lets Codex use desktop apps after your Mac locks, including remotely via Codex Mobile. The practical pairing is obvious: set a long-running goal at the end of the day, let the Mac lock, come back to a completed pull request.
The security model is worth understanding before enabling. OpenAI uses short-lived authorization tokens rather than persistent access. Codex scopes locked use to active, trusted computer use turns and includes safeguards. If the system detects local keyboard or mouse input, it immediately relocks and suspends automatic unlock until you manually unlock. Enable it in Codex Settings, then verify Screen Recording and Accessibility permissions in macOS System Settings.
Five more things in this release
The thread also shipped several team-oriented updates that matter for organizations running Codex at scale:
- Shared plugins: Teams can now distribute custom plugins across a workspace. Available now for Business users; enterprises can reach out for early access.
- Enhanced analytics: Businesses and enterprises get more detailed breakdowns of active users, credits, tokens, runs, user leaderboards, lines of code generated, and plugin usage, plus updates to the Analytics API.
- Advanced browser annotations: The in-app browser now lets you directly adjust page elements while leaving feedback, preview changes instantly, and batch comments for Codex, so designers can show exactly what they want changed without waiting for a full turn.
The bigger picture
This announcement is less about "Codex getting smarter" and more about strengthening the foundation for passing work context, delegating long tasks, verifying on-screen, and managing in teams. That's a meaningful distinction. The model hasn't changed. The scaffolding around it has.
Codex no longer has to be told what the developer is building, it can see it, and it no longer has to be re-prompted to keep going. Cursor, Claude Code, and others are converging on exactly this same model. The question worth watching isn't whether autonomous coding agents are useful, it's whether the output quality justifies the autonomy. That answer lives in your diff viewer, not the release notes.
How to get started
Goal mode is now live in the Codex app, VS Code and JetBrains extensions, and CLI version 0.128.0 or later. Update the Codex app to version 26.519 or later. On the CLI:
npm install -g @openai/codex # gets you 0.128.0+ for Goal Mode
# Enable Goal Mode if not already on
codex features enable goals
# Set your first goal
/goal increase test coverage in src/auth to 90%
# Check status, pause, or resume
/goal
/goal pause
/goal resumeIf /goal doesn't appear in your slash command list, you can also enable it manually in config.toml:
[features]
goals = trueAppshots are available across all plans on Mac. Enterprise access for Appshots is coming soon. The Appshots documentation covers the full list of supported apps and known limitations, and the Codex changelog has the full diff of everything that shipped in this release.