OpenAI Developers announced that dots, the always-on agents in ChatGPT, can now connect multiple local computers and choose which one to use for each task, for example building and testing an app on a dedicated machine while you keep coding on your laptop. The official docs were updated to match: each computer is approved separately from the ChatGPT desktop app on that machine, each thread uses one computer at a time, and independent tasks should run as separate threads per computer. The dot's cloud computer stays connected. Earlier docs said only one personal computer could be connected at a time. OpenAI has not published performance or reliability data.
Key Takeaways
- ✓Change: a dot goes from 1 connected personal computer to multiple local computers, plus its always-on cloud computer
- ✓Setup: open the dot profile in the ChatGPT desktop app on each machine, select Allow under Computers, then confirm Allow access; repeat these 3 steps per computer
- ✓Scheduling: each thread uses 1 computer at a time; for parallel independent tasks, ask the dot to start a separate thread on each computer
- ✓Permission isolation: the dot's local-computer access is separate from connecting a computer to Codex or enabling Work Sync; neither alone grants the dot access
- ✓The announcement post carries a 17-second demo video and drew 2,600+ likes and 130K+ views within about an hour; OpenAI published no performance data
Key Decision Metrics at a Glance
Turn your technical choice into a development budget
Compare 40 dev plans & simulate token costs vs $20/mo subscriptions
Project Links & Resources
Direct AccessIn-Depth Technical Analysis
Background and the pain point
Dots are OpenAI's always-on agents inside ChatGPT. Each dot has its own cloud computer and browser, and can also connect to your own computer to use local files, code and apps and to create local Work or Codex tasks. Until now the docs said you could connect only one personal computer at a time, so a developer who wanted a dot to run long builds and tests on a second machine had to swap the connection. On November 9 (US time), OpenAI Developers announced that limit is gone.
How it works
Per the updated official docs, the dot runs in the cloud and coordinates; local computers are execution targets it can use. Each computer is approved separately in its own ChatGPT desktop app, and the cloud computer stays connected. Each thread uses one computer at a time, so building on one machine and testing on another takes two threads, which the dot coordinates from the cloud. Connections persist between tasks; an offline machine shows Offline without losing access, and Revoke access removes a single machine. Dot access is separate from Codex connections and Work Sync.
Benchmarks
This is a feature update, not a model release, and OpenAI has published no performance, latency or reliability data. The announcement includes a 17-second vendor demo video and drew more than 2,600 likes and 130,000 views within about an hour. No independent test of multi-machine coordination exists yet.
Getting started
On each computer you want to add, open the ChatGPT desktop app, go to your dot's profile, find the machine under Computers, select Allow and confirm Allow access. A good setup is an always-on desktop or build box for heavy jobs and a laptop for interactive coding. Name the target computer in your request and ask the dot to open a separate thread per machine. Once approved, the dot can reach that computer's files, code and apps, so clear out sensitive credentials first and use Revoke access when you are done.
Benchmark side-by-side against alternatives, or calculate monthly token cost vs subscription break-even.
Discussion & Comments
0Sign in to join the discussion
Connect with AI developers to exchange benchmark insights.