Just now, OpenAI Codex has been launched on the ChatGPT mobile app, and all users can use it.
Everyone has been waiting for this for a long time.
This Friday, OpenAI announced the launch of the mobile version of Codex in the ChatGPT App. Both Android and iOS versions have started the preview, which is available to all users, including those on the free plan.
The main highlight of this update is to improve the synchronization rate. Now you can review code and assign tasks to AI anytime and anywhere. As soon as the news came out, the development and research communities immediately welcomed it. This is simply a qualitative change.
Currently, more than 4 million people use Codex every week. As AI agents take on more long - term and complex tasks, this new mobile - based collaboration model has emerged.
Codex is now integrated into the ChatGPT mobile application. You can stay updated on its latest operations, view Codex's discoveries, change directions, approve subsequent steps, or add new ideas. Meanwhile, Codex can complete work on your laptop, MacMini, devbox, or in a remote environment.
OpenAI said that Codex in the ChatGPT mobile app offers a full range of features, enabling you to use Codex efficiently to complete your work. When remotely connecting to any device running Codex, the app will load the real - time status of that environment, allowing you to smoothly handle various activity threads, approvals, plugins, and project contexts.
This is not just about remotely controlling a single task or assigning new tasks to a computer. You can handle all threads via your phone, view output results, approve commands, change models, or start new projects. Your files, credentials, permissions, and local settings are retained on the computer running Codex, while updates (including screenshots, terminal outputs, differences, test results, and approval information) are transmitted to your phone in real - time.
Codex uses a secure relay layer at its core, ensuring that trusted machines can access it on various devices without being directly exposed to the public Internet. Additionally, this relay layer allows you to keep your session state and context synchronized wherever you log in with ChatGPT.
With Codex, you can now:
Investigate bugs while waiting for coffee. Since Codex runs in your development environment, it can start checking relevant files, reproducing issues in the browser, running tests, and starting to fix them. If Codex needs your clarification or authorization to proceed, you can reply or approve directly on your phone. During the fixing process, you can view screenshots, terminal outputs, and test results, and check the final differences before returning to your computer.
Encounter a critical decision point during your commute. Before leaving for work, you ask Codex to handle a refactoring project that takes time to complete, planning to check the results when you reach the office. On the way, Codex finds two viable solutions and needs your instructions to continue. You can use your phone to weigh the pros and cons and choose a solution. By the time you reach the office, the project will have advanced as you expected.
Be better prepared for fast - paced customer conversations. After another meeting, you notice that a support issue is evolving across Slack, emails, documents, and browser tools, and you have a customer call coming up. At this time, you can use your phone to ask Codex to summarize the latest progress, mark key pending issues, and prepare a concise conversation briefing. If new information appears, you can ask Codex to update the summary before joining the conversation.
Turn your inspiration into action while it's still fresh. Whether you're having lunch, taking a walk, or listening to something inspiring, you can create a new thread on Codex via your phone or add it to ongoing work, and send your ideas to the AI for processing. In this way, tasks can start taking shape before you return to your desk, without making you completely miss the moment of inspiration.
Now you don't have to worry about AI "slacking off" anymore.
Currently, many teams are developing in hosted remote environments, which provide approved dependencies, credentials, security policies, and computing resources.
With the official release of remote SSH, Codex can directly connect to these environments. The desktop application will automatically detect the hosts in the SSH configuration and allow you to create projects and run threads on remote computers as if they were local.
After a successful connection, you can access these environments on authorized ChatGPT devices through the same secure relay infrastructure. This means you can start working on your desktop computer, control the execution via your phone, and keep long - running tasks going without being limited to a single device.
OpenAI said that they will also release several updates to further expand the ways in which teams can automate, customize, and manage Codex on a large scale:
Programmatic access tokens provide scoped credentials that can be issued directly from the ChatGPT workspace settings for CI pipelines, release workflows, and internal automation.
Hooks are now generally available and can be used to scan for secret information in prompts, run validators, record conversations, create memories, or customize Codex behavior for specific repositories and directories.
It supports using Codex in a HIPAA - compliant manner in local environments (CLI, IDE, App) to run ChatGPT Enterprise workspaces, enabling medical institutions to support patient care and operational workflows at a faster pace and with greater confidence.
The Codex feature in the ChatGPT mobile app is now in preview on iOS and Android platforms, covering all packages (including the free version and the Go version), and is available in all supported regions. To use it, you need to update the ChatGPT mobile app and the Codex app on macOS. OpenAI said that the mobile connection function for the Windows version of the Codex app will be launched later.
Reference content:
https://openai.com/index/work-with-codex-from-anywhere/
This article is from the WeChat official account "MachineHeart". Author: MachineHeart. Editor: MachineHeart Editorial Department. Republished by 36Kr with authorization.