What is OpenAI Dots? Cost, plans, and Slack for ChatGPT's new agents

OpenAI Dots are always-on agents inside ChatGPT, one per person to start, each with its own cloud computer and browser. OpenAI announced them at DevDay on September 29, 2026, and a dot keeps working between your conversations. One early tester, Dan McAteer, told The Next Web that his dot “proactively picked up on an invoice I needed to send for my freelance writing, pulled the relevant details from the email thread and drafted the invoice.”

The detail that matters most for teams is in Slack, where only a dot’s owner can direct it. Every OpenAI fact below comes from OpenAI’s announcement, safety post, help center, pricing pages, or the GPT-6 Astra system card, all read on launch day. We have not run a dot ourselves, so this is an explainer from OpenAI’s documents, not a hands-on review. We build Chickpea, a Slack agent, and it comes up by name only in the last section before the FAQ.

Key takeaways

• OpenAI Dots are always-on personal agents in ChatGPT, announced September 29, 2026. Each runs on GPT-6 Astra with its own cloud computer and browser, and reaches your apps through ChatGPT plugins.
• Your first dot is included with ChatGPT Pro, which costs $100 to $500 a month across three tiers, or with a Business Premium seat at $100 a month billed annually. OpenAI’s pages say Pro without naming a tier. Pro is excluded in the EEA, Switzerland, and the UK. Enterprise, Edu, and Healthcare get a beta an admin must turn on.
• Talking to your dot does not count toward ChatGPT usage limits. Tasks it starts in Codex or ChatGPT Work do. OpenAI has not published prices for extra dots or extra capacity.
• In Slack and Teams, only a dot’s owner can direct it. Colleagues can see what it posts, and their mentions do not start work.
• Consequential actions pass a separate check called Auto-review plus your own rules, and changing a password or moving money always comes back to you. OpenAI’s own line is that dots “can still make mistakes.”

What OpenAI Dots are

A dot is an agent tied to your ChatGPT account. You give it a goal, it works out the next steps, and it brings back work for review and decisions it needs from you.

It has its own cloud computer and browser, which you can open at any time to inspect its work, and it can use your laptop too once you connect it from the desktop app and confirm Allow access. That is off by default. Besides tasks you schedule, it runs background work OpenAI calls proactive research, reading your connected apps with read-only tools and saving notes for itself.

It reaches your apps through the same plugin permissions as ChatGPT, ChatGPT Work, and Codex, so an app you already connected is available to it within the access you granted. OpenAI puts the catalog at “over 4,000 apps.” It starts with your ChatGPT memory and builds its own, including from connected apps.

The model is GPT-6 Astra. According to the system card, dots run under a new time-budget setting that guides how long they work, and they often hand pieces of a job to subagents.

What OpenAI Dots cost and who can get one

Your first dot is included in ChatGPT Pro or a Business Premium seat at no extra cost. That is the whole price today, and the table shows what those plans cost as of September 29, 2026.

Prices are from OpenAI’s Pro tiers article and the in-app pricing dialog. Eligibility is from the getting started article. Rollout is gradual, and OpenAI says access “may take several days to reach your account.” Dots are not available to anyone under 18.

Conversations with your dot do not count toward your ChatGPT usage limits. When your dot starts or manages tasks in Codex or ChatGPT Work, those count as usual. Your plan also includes “an allowance for deeper work, with extended limits for the first month after launch.” OpenAI has not said how large that allowance is.

OpenAI says you will later be able to add more dots and raise each dot’s speed or the amount of work it takes on per month. It has not given prices for either.

Business Premium means the $100 Premium seat, not the $20 Standard seat. The regional exclusion applies to Pro only, so a Business Premium seat in Germany or the UK gets a dot.

Whether Pro 100 qualifies is still open. OpenAI’s pages say Pro without naming a tier, and we found no OpenAI page that settles it.

How to set up a dot

You create a dot on a computer, then talk to it anywhere. The steps from OpenAI’s help center:

1. Open the Mac or Windows desktop app, or ChatGPT in a desktop browser. You cannot create a dot on mobile, and mobile web is not supported.
2. Follow the onboarding, name your dot, and choose a character or a pet. Its handle is @yourname-dot, or @yourname-agentname once named.
3. Connect apps under Plugins. Each app shows its permissions before you connect it.
4. Optionally connect Slack or texting. Messaging channels are set up from desktop.
5. Optionally connect your own computer from the desktop app on that computer.

After that, the dot’s profile is where you manage it. In progress, Scheduled, and Completed show its work, and you can open its computer from there. Pause stops it until you resume. Reset deletes the dot with its conversations, saved memories, and scheduled tasks, and it is the only way to delete what the dot learned. Disconnecting an app stops new access but keeps what the dot already took from it, so connect the apps you would hand an assistant on day one, not everything at once.

At launch you can connect your personal email, but the dot cannot have its own address. You can start a voice call with it in ChatGPT, but it cannot call you. Texting is a limited beta for some Pro users in the US, run through a third-party provider, and not offered in Business or Enterprise.

In an Enterprise workspace, an owner turns on “Use dots (Beta)” first. Separate switches cover adding dots to Slack and Microsoft Teams, local computer access, and Custom Rules, and the last two are off by default. Workspace model settings do not apply to dots. OpenAI is also piloting specialist dots, each with its own identity, which a company sets up for one job, and says it is working with Microsoft to manage them through Agent 365. Neither has shipped.

How OpenAI Dots work in Slack and Teams

A dot in Slack is your agent posting where your team can see it. It takes instructions from you and nobody else. OpenAI’s admin article says it in three sentences: “Only a dot’s owner can direct it. Other people’s direct messages or mentions do not start work for that dot. When a dot posts in a channel, other channel members can see its messages.”

None of the launch-day coverage we read mentions that rule. You mention your dot, it answers, and it can post “with its own identity, where available.” A colleague who mentions your dot gets nothing. Context carries across channels, so a project you started in ChatGPT is still in context when you raise it in Slack.

Setup is per person. In an Enterprise workspace, both “Use dots (Beta)” and “Add dots to Slack and Microsoft Teams” must be on for the member. If the Slack workspace requires app approval, a Slack owner or app manager approves the install. Then each member completes their own dot’s Slack setup.

In practice, a channel where six people each set up a dot has six agents, each answering to one person and acting with that person’s app permissions.

The announcement and the Enterprise admin article both name Microsoft Teams, but the personal getting-started article mentions only Slack. If you are on Pro and want Teams, check the setup screen before you plan around it.

The safety questions

OpenAI put a separate reviewer between a dot and anything consequential, and it says the design can still fail. NBC News ran its launch story under the headline “OpenAI launches Dots AI agents amid safety questions.” The safety post and the privacy and safety FAQ describe the layers in order.

Plugin permissions decide which apps a dot can reach. Custom Rules let you set a behavior per action, from four choices: take action without asking, take action if pre-approved, ask before taking action, or hand off to you. “Pre-approved” means you asked for that action explicitly in your prompt.

Auto-review is a separate system that checks planned actions against your instructions, your Custom Rules, and OpenAI’s safety requirements before they run. Before an email goes out, it checks the recipient and the message. If it blocks a step, the dot can ask you, try a permitted alternative, hand the step back, or stop. OpenAI says the controls that enforce Auto-review sit outside anything a dot can change, and your approval cannot override its core requirements.

Some steps always come back to you. Changing a password and transferring money are hand-offs, which means the dot helps with the surrounding task and you finish the step. Permanently deleting data, installing software from an unrecognized source, and granting new security-sensitive access need your confirmation each time. Purchases with a card saved on a merchant’s site need approval, which you can give in advance for a specific purchase.

Proactive research is read-only, and OpenAI says it enforces that “in code.” Those tasks cannot send messages, change content in your apps, or control a browser or computer. Secure sign-in pauses the model while you type a password into a form, so the password never enters its context. That protection does not cover a password you paste into a chat or a document.

The GPT-6 Astra system card has a section on dots, and its numbers are OpenAI’s own evaluations. In a bulk test, OpenAI mixed 16,600 generated attack emails into 50,000 and recorded no successful attack. In an iterative test, where the attacker refined each email from the last response, 2,638 attempts also produced none. Across four simulated time budgets, from 4 minutes to a year, the model completed no actions that Auto-review prohibited. But after a warning told it an action was not allowed, it kept trying anyway in 17.4 percent of rollouts at maximum reasoning. The card also reports “a substantial decrease in chain-of-thought monitorability compared to previous models.” In plain terms, the model’s written reasoning is a weaker signal of misbehavior than it was in earlier models. These are OpenAI’s tests of its own product, not an outside audit.

Your data follows the plan you are on. OpenAI does not train on Business, Enterprise, or Edu content by default. On personal plans, the “Improve the model for everyone” setting decides, and OpenAI says it does not train directly on proactive research or its notes. Human review can still happen in safety cases with that setting off. You cannot view, edit, or delete individual dot memories. Deleting the dot is the only way to clear them, and that is the rule to weigh before you connect a work inbox.

Where Dots is the better choice

For your own work, across your own accounts, a dot is the better fit, and a shared agent is the wrong one.

It works in your inbox, calendar, and documents with memory from your ChatGPT history, and those are accounts you would not hand to a bot the whole channel can mention. It comes with a cloud desktop and browser, plus your laptop if you allow it, with nothing to set up. It reaches you in ChatGPT on desktop, web, and mobile, in Slack, in Teams where available, and by voice in ChatGPT.

You deploy nothing, and the apps you already connected in ChatGPT carry over with the permissions you gave them. If you pay for Business Premium, or for a Pro tier that qualifies, the first dot costs nothing extra.

When a team needs a shared agent instead

If the job belongs to a channel rather than a person, such as triaging a support queue, taking intake requests, or posting the weekly metrics, you want an agent the whole channel can direct, with accounts that belong to the job. By OpenAI’s own rule a dot cannot be that, and OpenAI sells two products that can.

@ChatGPT in Slack and Teams is on Business and Enterprise. An admin configures its instructions and company connections, it runs through a service account rather than any one person’s, and teammates can use it without their own ChatGPT license. Workspace agents, through the ChatGPT Agents app, give each agent its own Slack handle, let it answer in channels you choose, and run it on a schedule. They need a Business, Edu, or Enterprise plan and a paid Slack plan. If your company already standardizes on ChatGPT Business or Enterprise, start there.

The other route is named agents on infrastructure your team owns, which is what we build. In Chickpea, each Agent is one teammate with its own Slack handle, instructions, model, memory, and connections. Every member of a channel it is published to can mention it, and the work runs with that Agent’s accounts, not the asker’s. Consequential actions such as sending, deleting, publishing, and bulk changes wait for a person unless the Agent’s saved instructions allow that class of action. A check outside the model also refuses any write the current Slack request did not ask for. It is Apache 2.0 software in your own Cloudflare account or on a Node host, with your own model keys and no seats.

You give up things a dot includes. You run the deployment yourself. Named handles need a paid Slack plan and a workspace setting that lets members create user groups, per the requirements page. It works in Slack only, with no Teams, texting, or voice. An Agent has no desktop of its own, and a web browser means connecting a Browserbase project. The personal-versus-shared split is the same one we walked through for OpenClaw teams.

The choice is whose work it is. If it is yours, set up a dot. If it is the team’s and your company is on ChatGPT Business or Enterprise, start with @ChatGPT or workspace agents. If it is the team’s and you want the agents and model keys on your own account, the Chickpea install guide starts with one deploy to your Cloudflare account.

FAQ

What are ChatGPT dots?

Dots are always-on agents in ChatGPT that OpenAI announced on September 29, 2026. Each one runs on GPT-6 Astra with its own cloud computer and browser, connects to your apps through ChatGPT plugins, and keeps working between conversations.

How do I use ChatGPT dots?

Create your dot in the ChatGPT desktop app or in ChatGPT on a desktop browser, name it, and connect the apps it may use. After setup you can message it in the ChatGPT mobile app, and connect Slack or texting from desktop.

How much does OpenAI Dots cost?

The first dot is included with ChatGPT Pro, which costs $100, $200, or $500 a month, or with a Business Premium seat at $100 a month billed annually or $125 monthly. OpenAI’s pages say Pro without naming a tier, and it has not published prices for additional dots or extra capacity.

Is OpenAI Dots available in the UK and Europe?

On Business Premium, yes, in all supported ChatGPT regions. On Pro, no. The rollout excludes the European Economic Area, Switzerland, and the UK.

Can other people in a Slack channel use my dot?

No. OpenAI’s admin article says only a dot’s owner can direct it, and other people’s mentions or direct messages do not start work, though channel members can see what it posts. For an agent the whole channel can use, OpenAI offers @ChatGPT in Slack and workspace agents on Business and Enterprise.

Can I have more than one dot?

Not yet. You start with one primary dot, and OpenAI says you will be able to add more later, at a price it has not announced.

Is OpenAI Dots safe to connect to my accounts?

OpenAI routes consequential actions through Auto-review and your Custom Rules, always hands password changes and money transfers back to you, and keeps proactive research read-only. It also says these protections reduce prompt-injection risk “but they do not eliminate it,” so review anything consequential before it goes out.

Is “dots AI” the same as OpenAI Dots?

Not always. Several unrelated companies and models use the name Dots or Dot, so check that a result is about OpenAI before you rely on it. OpenAI’s product lives in ChatGPT and is documented at openai.com and help.openai.com.

Sources

OpenAI, product: Introducing dots, How we build safety, security, and privacy into dots, DevDay 2026 recap (all September 29, 2026). OpenAI, help center: Getting started with your dot, Dots privacy, security, and safety FAQs, Manage dots in ChatGPT workspaces, About ChatGPT Pro tiers, Using @ChatGPT in Slack and Microsoft Teams, ChatGPT Agents app in Slack (all checked September 29, 2026). Business seat prices from the ChatGPT pricing dialog, checked September 29, 2026. OpenAI, safety: GPT-6 Astra system card, section 12, Dots (checked September 29, 2026). Press: The Next Web, OpenAI launches dots, NBC News, OpenAI launches Dots AI agents amid safety questions (both September 29, 2026). Chickpea: what an Agent is, handles and channel grants, authority and confirmation, models and providers, browser, requirements and costs (verified September 4 to 22, 2026), pricing.

Facts checked September 29, 2026, launch day. OpenAI’s help center was updating hourly, and the Pro tier question and Teams for individual users are the items most likely to change. We re-check the register behind this page at 14 days and again at 30, 60, and 90.