OpenAI's Dots Agents Can Now Control Multiple Computers at Once
ChatGPT dots can now be linked to several local machines at once, with per-task routing so a build can run on one Mac while you code on another.
- OpenAI dots can now connect to multiple local computers, with per-task machine selection via the docs.
- Each task still uses one computer at a time; start separate threads for parallel work across machines.
- Enables a dedicated Mac to run Xcode builds while your laptop stays free for coding.
- Useful for cross-SDK bug reproduction and UI regression sweeps without hijacking your main screen.
- Available to Pro and Business Premium subscribers; requires the ChatGPT desktop app running on each linked machine.
- Local access is still opt-in per machine, revocable, and separate from Codex or Work Sync permissions.
OpenAI has expanded local-computer access for its dot agents, allowing one dot to connect to several computers and letting the user choose which machine handles each task. Developers can send a long build, test suite, or browser-driven workflow to a secondary machine while keeping their primary computer available.
A dot is a persistent AI agent powered by GPT-6 Astra. OpenAI provisions each one with a dedicated Linux computer and Chrome browser that persist between conversations. That cloud environment remains the default workspace, while access to local computers requires explicit approval.
One dot, several computers
The previous documentation limited each dot to one linked desktop. OpenAI’s updated Computers and apps documentation now describes support for multiple registered machines:
- Open the dot’s profile in the ChatGPT desktop app on the computer you want to add.
- Find that machine under Computers and select Allow.
- Confirm with Allow access, then repeat the process on any additional computers.
Each task uses one computer at a time, with the cloud machine and approved local computers available as targets. Concurrent jobs require separate threads, each assigned to its own machine. Messages sent to the same dot from a phone or another device can also direct work to an approved computer.
Where a second machine pays off
Computer-use jobs can occupy the cursor and foreground applications for extended periods. Routing them to spare hardware keeps the main development environment free and supports several practical workflows:
- iOS builds on a dedicated Mac. The dot can open Xcode, apply a change, run the build, and return screenshots while development continues elsewhere.
- Cross-environment reproduction. A developer can run the same test on a current setup and on a machine pinned to an older SDK, then compare the results.
- UI regression sweeps. The dot can navigate a flow on a secondary computer, record failures, and capture screenshots without taking over the primary display.
- Long-running test suites. Integration tests, browser automation, and repeated build loops can run on hardware that would otherwise sit idle.
Permissions stay machine-specific
OpenAI keeps every local connection explicit and revocable. The operating requirements and permission boundaries are:
| Area | Behavior |
|---|---|
| Availability | The computer must remain online with the ChatGPT desktop app open. An Offline status makes it unavailable without removing its registration. |
| Approval | Every computer requires a separate approval from the dot’s profile. |
| Revocation | Access can be removed for an individual machine without disconnecting the others. |
| Browser sessions | The dot’s cloud browser maintains its own sessions and does not inherit logins from the user’s personal browser. |
| Related services | Connecting a computer to Codex or enabling Work Sync does not grant that computer to a dot. Each feature has separate permissions. |
Machine choice cannot rescue a failed task
Independent reporting based on OpenAI’s dots system card says the agents fail nearly one in five long-running tasks. Additional computers provide more execution capacity, but they do not improve the agent’s ability to complete a multi-step workflow correctly.
A single task also remains bound to one computer. The dot cannot divide one job across two local machines or coordinate both within the same thread. Parallel execution requires separate threads, and developers still need to inspect outputs, retry failed runs, and verify code changes.
Access follows the existing dot tiers
OpenAI introduced dots at DevDay 2026 for Pro and Business Premium subscribers in eligible markets. The company also offers a $500-per-month Pro 500 plan with roughly 25 times the ChatGPT Plus usage allowance and exclusive access to Astra Ultrafast speed.
Multi-computer support is included with the existing dots feature set. There is no separate add-on, although every local machine needs the ChatGPT desktop app, an explicit approval, and an active connection while the dot is using it.
Spare hardware becomes a runner pool
The updated architecture resembles a small pool of developer-controlled runners. The dot’s cloud computer handles isolated browsing and research, while approved local machines provide access to installed SDKs, native development tools, test environments, and authenticated applications.
A developer with a work laptop, personal desktop, and spare Mac can now route agent tasks across those systems without operating a separate orchestration service. The practical gain is straightforward: long-running agent work can move to the machine best suited to execute it, leaving the primary development environment available.
Comments
