OpenAI lets dot connect to multiple local computers

If you have a spare Mac or an old PC sitting on your desk gathering dust, OpenAI just found a very specific job for it.

OpenAI updated its persistent AI assistant, dot, allowing users to connect multiple local machines to the same dot instance and assign tasks to specific computers on demand.

First introduced at DevDay in late September, dot lives in the cloud with its own virtual machine and browser, working in the background even when your laptop is closed. While it could previously link to a single local computer to access local files and apps, multi-machine support lets users offload heavier builds and messy environments entirely to dedicated hardware.

Offloading builds and preserving your cursor

OpenAI's developer team highlighted a few clear use cases that will sound instantly familiar to engineers.

For example, while coding an iOS app, you can tell dot to build and test your changes inside Xcode on a secondary Mac and message back a screenshot of the result. For bug hunting, dot can run the same test on both your main laptop and an older machine loaded with legacy SDKs to compare results, without messing up your daily environment.

It is also a lifesaver for UI testing: dot can drive the mouse and click through user flows on a second machine to spot breakages, while your primary screen and keyboard remain entirely yours to keep writing fixes.

Under the hood, dot kicks off Codex or ChatGPT Work local sessions on the chosen machine. Each thread runs on one designated computer at a time; running jobs across multiple machines requires opening separate threads. Tasks already in progress will not migrate if you change the machine selection mid-run.

To hook it up, users must open the ChatGPT desktop app on each target computer, head to dot's profile under Computers, and grant access individually. The connected machine must stay awake, online, and have ChatGPT running. The rollout is currently expanding across ChatGPT Pro tiers and Business Premium (18+, excluding the EEA, UK, and Switzerland). Enterprise, Edu, and Healthcare tiers require admin opt-in today, turning on by default on November 18 unless disabled.

Why it matters

In our view, this update targets one of the most annoying friction points of local AI agents: environmental and cursor conflict.

Pure cloud agents are great because they run while you sleep, but they lack your proprietary local SDKs, private credentials, and local build pipelines. Meanwhile, running local UI automation on your primary machine often hijacks your mouse and pegs your CPU. By letting a cloud orchestrator dispatch work across your secondary hardware, OpenAI essentially turns your spare machines into private build nodes.

This also sharpens the contrast with rivals. Anthropic's Claude Duet relies heavily on cloud VMs and over 6,000 connectors, keeping off your local desktop entirely unless you touch Claude Code. xAI's Grok Bot does allow command execution on local machines with optional IP routing through your hardware, but dot's integration with Codex and ChatGPT Work makes it feel much more tailored to continuous development workflows.

For developers juggling cross-platform tests or heavy local builds, it is well worth setting up. Turning that secondary machine into an autonomous testing sandbox is a clever, high-utility move.