Codex Remote is now generally available on all ChatGPT plans
Codex Remote reached GA on 25 June 2026, letting every paid ChatGPT subscriber start, steer, and approve Codex work from their phone via secure QR pairing.
On 25 June 2026, OpenAI made Codex Remote generally available across every paid ChatGPT plan. That means Plus, Pro, Business, Enterprise, and Education subscribers can now start, monitor, and steer long-running Codex coding sessions directly from the ChatGPT mobile app, with the actual work executing on a connected Mac or Windows host.
More than 4 million people use Codex every week. The GA release is OpenAI acknowledging that agents working across hours rather than seconds need a different kind of human oversight, one that does not require you to stay at your desk.
What Codex Remote actually does
The framing that matters here is not “Codex on your phone.” Your phone is not doing the heavy lifting. The Codex agent runs on a paired host, such as your MacBook, a Windows workstation, or a cloud machine. Your phone becomes the control surface.
From the ChatGPT mobile app you can:
- Start a new Codex thread on a connected host
- Continue or redirect work already in progress
- Review diffs, terminal output, test results, and screenshots
- Approve actions the agent is waiting on
- Answer questions Codex raises mid-task
- Switch between multiple paired hosts
The practical shift is that a long-running refactor or migration no longer stalls because you stepped away. You can answer a question or approve the next step from wherever you are, and the agent keeps moving.
How pairing works
Each mobile device and host pair through a QR code displayed in the Codex desktop app. Scan it once from the ChatGPT mobile app and the two are linked with authenticated one-to-one pairing.
Under the hood, this uses the Noise Protocol Framework. Since version 0.141.0 of the Codex app (released 18 June 2026), remote executors communicate over Noise protocol relay channels. The relay uses Diffie-Hellman key exchange over Curve25519 and AEAD symmetric encryption, so the OpenAI relay infrastructure sees only encrypted ciphertext. It cannot inspect your commands or outputs.
The trust model is account-based. There is no separate relay credential. Your ChatGPT account authorises the connection, which is worth noting if the host holds production credentials.
A few practical points on connections:
- Pairings active since 8 June 2026 carry over automatically. Older, inactive connections need to pair again.
- Signing out of the mobile app disables Remote Control but does not remove existing pairings.
- On Mac laptops, the lid must stay open with power connected, or an external display must be attached, for the host to remain reachable.
- Windows hosts running Computer Use must stay unlocked.
- Update both the ChatGPT mobile app and the Codex desktop app to the latest versions before connecting.
For headless or terminal-only setups, the CLI also offers a manual pairing path that generates a code you enter in the mobile app rather than scanning a QR code.
The DigitalOcean plugin removes the local machine requirement
Alongside the GA release, OpenAI shipped a DigitalOcean Droplet Workspace plugin for Codex. It provisions a Droplet from a Codex-ready image, generates and configures SSH access, and connects the new machine to your Codex app as a remote workspace. No manual API tokens or doctl CLI required.
Tasks running on a Droplet are not affected by your laptop going to sleep or a network interruption on your local machine. The Droplet connects through the same Noise relay as any other paired host, so you monitor and direct it from your phone exactly the same way.
This matters for developers who do not want to leave a local machine running for hours. The cloud workspace becomes the host, and your phone and laptop are just control surfaces.
Remote SSH is also now generally available
Remote SSH reached general availability alongside Codex Remote. The Codex desktop app automatically detects hosts from your SSH configuration and lets you create projects and run threads inside remote machines as if they were local. Those environments can then be accessed across your authorised ChatGPT devices through the same relay infrastructure.
For teams that already develop inside managed remote environments with approved dependencies, security policies, and fixed compute, this means Codex can run inside those environments without workarounds.
What this means for you
If you run long tasks with Codex: The approval bottleneck changes meaningfully. Multi-hour tasks have always had a structural problem: the agent needs human input at unpredictable moments, but keeping a developer at their desk is costly. Remote GA does not solve every interruption, but it removes the constraint of being physically present at the host machine.
If you work across time zones or irregular hours: A thought at lunch or on a commute can become a task in motion before you return to your desk. Start a thread from your phone, point it at the right host, and the work begins.
If you are on Enterprise or Education: Codex Remote GA is available in those workspaces. The same mobile control surface applies, within your workspace’s existing policies.
If you want a cloud-resident workflow: The DigitalOcean plugin means you can provision, pair, and monitor a remote coding environment entirely without a local machine as the host. Combined with OpenAI’s June 2026 acquisition of Ona (formerly Gitpod), the direction of travel for Codex is clear: agents that persist and continue working independently of any single developer machine.
Anthropic shipped comparable remote-access functionality for Claude Code in February 2026, so this space is moving quickly. The GA release brings OpenAI’s implementation to every paid plan with a security model that has been designed for the relay to remain blind to your actual work.
If you are already using Codex, update your apps and pair your phone. If you have been waiting for a reason to try it, a mobile control surface with proper cryptographic pairing is a reasonable place to start.