What are OpenAI Dots? Features and access
OpenAI Dots are always-on agents in ChatGPT. Learn how to get a dot, what it can do, launch pricing and limits, and how its approvals work.
OpenAI Dots are always-on personal agents in ChatGPT. A dot has its own cloud computer, works across apps you connect, keeps context between conversations, and can take on continuing work. OpenAI introduced Dots on September 29, 2026. This page covers what the launch actually includes and where a separate AI worker fits.
What can a dot do?
A dot is powered by GPT-6 Astra. It can browse, create files, use connected apps and continue work after a conversation ends. You can give it several projects and inspect its computer and activity. OpenAI gives examples that span tested code fixes, launch materials, updated research figures and sales proposals. These are examples of intended uses, not a guarantee that every dot can perform every task reliably.
Dots can also run recurring tasks you schedule. Their proactive research checks connected sources with read-only tools while you are away. That distinction matters: OpenAI says proactive research itself cannot send messages, change content or control a browser or computer. Follow-up actions use the normal approvals and safeguards. See OpenAI's safety explanation for the current details.
| Capability | At launch |
|---|---|
| Primary dot | One dot you name and personalize |
| Computer | Its own cloud computer and browser; connecting your local computer is optional |
| Channels | ChatGPT; Slack and Teams messaging where configured; texting is a limited beta |
| Apps | Connected through ChatGPT's plugin controls, subject to available apps and permissions |
| Work while away | Scheduled tasks and read-only proactive research; other actions follow approval rules |
| More dots | OpenAI says teams of dots and additional dots are future plans |
Who can get OpenAI Dots?
OpenAI says the rollout begins with Pro and Business Premium users in eligible markets. Enterprise workspaces, including Edu and Healthcare, can try a beta when an admin enables it. The getting started guide says Pro rollout excludes the European Economic Area, Switzerland and the UK at launch, while Business Premium is available across supported ChatGPT regions. Access is gradual and may take several days.
Create your first dot on ChatGPT desktop web or in the desktop app. Once set up, you can message it in the mobile app when mobile access is available. The help article says you cannot create a dot on mobile or use it on mobile web at launch.
What do OpenAI Dots cost at launch?
The first dot is included in eligible Pro and Business Premium plans. OpenAI says deeper work will have an allowance, but its getting started guide says dot usage will not count toward eligible Pro, Business and Enterprise plan allowances for the first month after launch. OpenAI plans to publish terms for each plan after that period. Conversations with a dot do not count toward ordinary ChatGPT message limits; Codex or ChatGPT Work tasks started by a dot still use those products' usual limits. Check current plan terms before planning a high-volume workflow.
What controls do you have?
You choose which apps to connect, can follow work in Activity View and can pause the dot. Custom Rules can allow, require approval for, or block specific actions within OpenAI's mandatory safety requirements. Sensitive actions can require confirmation or a handoff back to you. A dot's local computer access starts off, and you must choose to connect it.
These controls belong to Dots. They do not automatically scope an unrelated worker running elsewhere. Review the permissions in each system when a workflow crosses them.
How is a dot different from an ordinary ChatGPT chat?
The difference is continuing responsibility. In a normal chat, you return and ask the next question. OpenAI describes a dot as carrying project context, making progress between conversations, reporting back and raising decisions that need you. It has a computer you can inspect and an Activity View for work in progress. That does not mean it may act without bounds: connected app permissions, Custom Rules, required approvals and safety checks still govern its actions.
A useful first task has a small outcome you can verify. Give your dot an existing project and ask it to gather the current status from apps you have connected, identify three open decisions and prepare a draft for review. Watch which sources it uses and whether it asks for access or approval. Then try a recurring check, such as a morning calendar review, and inspect how it appears under Scheduled. This is a more informative trial than asking the dot to "run everything" before you know its reach.
| First-week check | What to look for |
|---|---|
| Connected sources | Can you see which apps are connected, and did the dot use only the ones needed? |
| Activity | Can you follow the work and correct it while it is in progress? |
| Review | Does it distinguish a draft from an action already taken? |
| Schedule | Does the recurring check run at the time you chose, and can you pause it? |
| Cost and limits | Which work is included during the launch period, and which tasks use separate Codex or ChatGPT Work limits? |
Do not treat "available 24/7" as a promise of unlimited complex execution. The first-month usage policy is temporary, and separate Codex or ChatGPT Work tasks keep their own limits. Recheck OpenAI's terms before turning a high-volume process into a standing task.
Where does WorkerKit fit?
A dot can manage broad, personal projects. A WorkerKit worker is built around one repeatable job: its own written instruction, app grants, schedule, model, budget and run receipt. That separation can help when a daily job should run independently of your dot's conversation and report only what changed.
The launch-day handoff we can describe concretely is email. OpenAI says a dot can use a connected personal email account. After you connect a sending mailbox in WorkerKit, a worker can email its report to the account your dot reads. Ask the dot to review that message. This is a one-way report delivery. It does not grant the dot WorkerKit fleet tools. The OpenAI Dots guide explains the fit, and the setup page walks through the email route.
For example, a meeting prep kit can run on a schedule, use only the accounts approved for that worker and leave a receipt. You could send its completed briefing to a connected mailbox, then ask the dot to compare it with the live project context and prepare questions for the meeting. The dot's synthesis and the worker's scheduled execution are distinct steps, so you can inspect which system produced each claim. The worker schedule and the dot's own scheduled tasks are configured separately.
OpenAI describes a plugin ecosystem for Dots. That makes a future direct WorkerKit plugin plausible, but it does not by itself verify one. We will update the setup procedure when a direct path is tested.
Dots, Muse and Grok Bot
Dots arrive in the same new category as Meta Muse and Grok Bot: agents with a computer, connected services and work that continues after you send the first message. The products differ in their ecosystems, channels and controls. Our Dots vs Muse vs Grok Bot comparison separates what each vendor has announced from what still needs hands-on verification.
FAQ
Are OpenAI Dots available to ChatGPT Plus users?
OpenAI's launch-day rollout names Pro and Business Premium, plus an Enterprise beta enabled by an admin. It says access will expand. The announcement does not list Plus for the initial rollout, so check OpenAI's current eligibility page before upgrading solely for Dots.
Can I create a dot on my phone?
Not at launch. OpenAI's getting started guide says to create a dot on desktop web or in the desktop app. After that, the mobile app can message it when mobile access is available. Mobile web is not supported at launch.
Does a dot have its own email address?
No. The help article says you can connect a personal email account but cannot give a dot its own standalone email address at launch. That is why the WorkerKit handoff uses a mailbox you control, not an address invented for the dot.
Can a dot make changes while I am away?
It can work on ongoing tasks, but OpenAI distinguishes scheduled tasks and read-only proactive research. Actions that affect accounts or share information follow its rules, review and approval system. Check the Activity View and read consequential results before relying on them.
Can I connect WorkerKit directly to a dot?
OpenAI documents plugins for Dots, but a direct WorkerKit plugin or MCP connection in Dots has not been verified here. The Dots and WorkerKit setup guide uses connected email for report delivery. It leaves fleet control in WorkerKit or another tested client.
Sources: OpenAI's Dots announcement, OpenAI's getting started guide, and OpenAI's safety overview. Product terms and availability can change after launch.