What Are OpenAI’s Dots? The Always-On Agent Built to Take On Meta’s Muse

OpenAI dots are always-on agents inside ChatGPT that keep working on a goal between your conversations. OpenAI launched them at DevDay on September 29, 2026, three weeks after Meta released Muse, its own personal AI agent.

A dot is different from a chat. You give it a goal and a set of permissions, and it keeps making progress in the background. It works through the apps you connect, on its own cloud computer, and comes back to you when it needs a decision.

The short version: dots are available now to ChatGPT Pro and Business Premium users, with Enterprise, Edu and Healthcare in a beta that admins must turn on. Your first dot costs nothing extra, and usage does not count against your plan for the first month. OpenAI has not said what comes after that.

This explainer covers what a dot does, who can get one, how you control it, and how it lines up against Meta’s Muse. It is a snapshot as of September 30, the day after launch.

What OpenAI dots actually do

OpenAI’s dots announcement describes them as always-on agents that learn your preferences and keep working toward a goal. Its help center calls a dot "an always-on agent in ChatGPT that can take on ongoing responsibility."

In practice, a dot has three things a normal chat does not.

• A standing goal. You set it once, and the dot keeps working on it across days.
• Its own workspace. Each dot has its own cloud computer and browser, plus the apps you connect.
• Memory. Dots receive memories from ChatGPT and create their own as they work.

OpenAI says dots can connect to more than 4,000 apps. They can also use your laptop, but only if you allow it, and local computer access is off by default.

Dots run on GPT-6 Astra, OpenAI’s top model. That matters for cost and for quality. It is the same model family we looked at in what GPT-6 Astra actually shipped, and the one whose next version OpenAI just pulled over safety concerns.

Where you use a dot

You create a dot in ChatGPT on the desktop app or the web. OpenAI’s getting started guide says you can check on a dot from mobile, but you cannot create one there yet.

Dots can also join Slack and Microsoft Teams, where they post on their own. OpenAI says texting is coming soon. At launch, a dot cannot call you.

During setup, you give the dot a goal and decide what it can do on its own. You can also give it a name and an avatar. That sounds cosmetic, but it helps when a dot starts posting in a shared Slack channel.

Who can get OpenAI dots

Availability is narrower than the launch coverage suggests.

OpenAI says the rollout is gradual and may take several days to reach every account. If you cannot find dots today, that alone does not mean your plan is excluded.

For most small nonprofits and associations, the plan question comes first. A standard Business plan is not listed. Dots need Business Premium, Pro, or an Enterprise-tier plan.

What OpenAI dots cost

The first dot comes at no extra cost on eligible plans. OpenAI’s help center adds that dot usage will not count toward your plan’s allowance for the first month.

After that, OpenAI has not published pricing for dot usage or for additional dots. Budget for the trial month only, and check OpenAI’s pricing again before it ends.

OpenAI also announced a new Pro 500 plan at DevDay. Its DevDay recap describes it as 25 times the Plus allowance, with access to the new Ultrafast speed tier. Fortune reports the price as $500 a month. Most small teams will not need it to try dots.

How you control a dot

Controls matter more for an agent that acts while you are away. OpenAI’s dot controls page describes three layers.

Automatic review. Before a dot takes an action that could affect your accounts or share information, OpenAI says an automatic review checks it against your instructions, permissions, custom rules and built-in safety requirements. Some actions, such as changing a password, stay with you.

Custom rules. Under Settings > Personalization > Custom rules, you choose how the dot handles a type of action. The four options are: act without asking, act when you say so, ask first, or hand it off to you.

Activity view. A dot’s profile has an Activity tab showing progress, files, results and background work. That is where you step in when a dot is waiting on an approval or a sign-in.

OpenAI is plain about the limits. It says custom rules are instructions the dot tries to follow, and it can make mistakes. Rules do not grant access, override safety requirements or remove required confirmations.

To clear a dot’s memories, you reset it, which deletes the dot entirely.

Admin settings for Enterprise and Edu

For organizations on Enterprise or Edu, OpenAI’s workspace admin guide lists four permissions.

• Use dots (Beta): off by default for Enterprise.
• Add dots to Slack and Microsoft Teams: lets dots join and post in those tools.
• Allow local computer access: lets a dot use a member’s files and run commands on their computer.
• Use custom rules for dots: lets members write their own rules.

Two details are easy to miss. First, turning on dots does not open every app; plugin controls and app permissions still decide what a dot can reach. Second, OpenAI says a role that grants dots access can still allow it even if another role has it turned off. Check every custom role, not just the workspace default.

Enterprise model controls also do not apply to dots. If your organization limits which models staff can use, dots sit outside that setting.

OpenAI dots vs Meta Muse

Meta launched Muse on September 8. Its announcement calls it a "secure, private personal AI agent" that proactively helps people meet their goals. On the surface, the two products do the same job.

The difference is the audience. Muse is a consumer product first, with shopping and travel connections. Dots sit inside a paid ChatGPT plan, with Slack, Teams and enterprise admin controls from day one.

Meta is moving toward business, too. TechCrunch reports that Meta announced Muse for Small Business on September 29, the same day as DevDay. We could not confirm the details on a Meta page yet.

We covered the model behind Muse in our Muse Spark 1.1 explainer in July.

Should your organization try OpenAI dots?

For most agencies and nonprofits, dots are worth a small test if you already pay for an eligible plan. The trial month costs nothing extra.

Pick one repeating job with a clear finish line, such as a weekly status summary from Slack and your project tracker. Start with "ask first" rules on anything that sends messages or changes data. Leave local computer access off.

If you run ChatGPT Enterprise or Edu, decide before staff ask. Dots are off by default, so nothing changes until an admin acts. Our overview of ChatGPT Work covers the related agent. OpenAI’s admin settings for cloud browser and cloud computer use apply to both.

Frequently Asked Questions