The interesting part of this story isn’t that OpenAI is testing Dots on ChatGPT Plus. It’s that an agent built to work without supervision is already being pushed to the $20 tier, while the company still hasn’t said what it costs users to run one.
OpenAI announced the test through ChatGPT engineering lead Thibault “Tibo” Sottiaux, who posted on X that a small number of Plus users are getting access. He also said the team is seeing “a lot more positives” about how well the proactivity works after several rounds of fixes. Dots also arrive for all ChatGPT Business Standard users starting today.
Dots are always-on agents. They work on tasks without prompting, keep making progress when you’re away, and reach more than 4,000 apps through OpenAI’s plugin ecosystem. They can also connect to your PC and control it alongside you, and you can open a Dot’s “computer” to see what it’s doing. Until now, you needed ChatGPT Pro or higher to use them.
The pricing details are the catch. On Pro and Business Premium, chats with Dots don’t count against normal ChatGPT usage limits, unless the work is complex. Whether Plus and Business Standard get the same treatment is unknown. That matters a lot. Agents that run in the background burn through compute, and a Plus user with a tight cap could hit it fast. And OpenAI hasn’t said if the Plus access will stay.
Why push it now? Because everyone is racing to put agents in front of ordinary paying users. Anthropic has been building computer-use features into Claude, Google keeps pulling Gemini deeper into Android and Workspace, and Microsoft is tying Copilot to Windows. OpenAI’s edge here is the plugin catalog and a huge installed base of Plus subscribers. Moving Dots to $20 turns a premium feature into a mass-market one, and it gives OpenAI far more feedback than a Pro-only rollout would.
The product is also getting more capable. Dots launched on GPT-6 Astra and can now run on GPT-6.1 Terra. OpenAI also added iOS widgets, Shortcuts integration, iMessage and RCS support, and notification replies on iOS. Those are the sort of features that decide whether people actually use an agent day to day, instead of trying it once and forgetting it.
So does this matter? Yes, with a caveat. A small test is not a launch, and the missing limit information is a real gap. If you’re a Plus subscriber, you may see Dots today. If you’re building or buying agent tools, watch how OpenAI settles the usage rules. That will tell you more than any feature list.



