OpenAI Dots (everything we know)
OpenAI has launched Dots, a new class of always-on AI agents that can keep working on a user’s behalf without waiting for the next prompt. The company says Dots can take on projects, use their own cloud computer and browser, connect to more than 4,000 apps, and keep working toward a goal around the clock.
That makes Dots much more than another chatbot feature that OpenAI is famous for, as OpenAI is now moving ChatGPT into territory already being claimed by Meta’s Muse and xAI’s Grok Bot, with each company trying to turn AI from something you talk to into something that actually takes work off your hands.
OpenAI just raised the pressure on the entire AI agent race. Dots look like the company’s direct answer to Muse and Grok Bot, but OpenAI is betting that its massive ChatGPT ecosystem, deep app connections and GPT-6 Astra model can make its version harder to ignore.
 |
| Credit: OpenAI |
Dots are built to keep working after you leave
OpenAI describes Dots as “always-on frontier intelligence agents” powered by GPT-6 Astra. Each Dot gets its own cloud computer, its own browser and access to apps the user connects.
The difference is what happens after the user ends the conversation, the Dots keep on working.
A Dot can take a project and continue working on it, even as the user gives it other jobs.
OpenAI says users do not need to keep directing every step. The agent can investigate a task, work through several stages and return with something ready for review.
OpenAI’s examples include a software developer whose Dot watches customer feedback, identifies smaller fixes, builds and tests them, then prepares pull requests with videos for review.
- A sales lead can have a Dot track changing customer requirements, build a proof of concept and update a proposal.
- A content creator can hand over interview transcripts and have the agent identify clips, prepare show notes and draft social posts.
Those examples point to the product OpenAI is actually selling: not faster answers, but delegated work and it can actually order food for you:
 |
| Credit: OpenAI |
Your Dot is supposed to learn how you work
OpenAI says the agent can learn a user’s preferences, standards and expectations from ongoing interaction.
The company wants the relationship to feel less like opening a new AI chat every time something needs doing and more like having a persistent digital worker that already knows the context.
Dots can be reached through ChatGPT on desktop, web and mobile. OpenAI is also bringing them into Slack and Microsoft Teams, with texting planned for later. The company says context carries across those channels, so users can start a project in ChatGPT and continue working with the same Dot inside a team environment.
That cross-platform approach could matter as much as the model itself. An agent that lives only inside one chat window is useful. An agent that follows a project into the places where the work actually happens becomes much harder to treat as a simple chatbot.
OpenAI is not hiding the agent from you
Giving an AI access to apps, accounts and computers creates an obvious problem: the more freedom the agent gets, the more damage a bad action can cause.
OpenAI says Dots run on separate cloud computers, keeping a user’s computer and its files apart unless the user chooses to connect them. Users control which apps a Dot can access and can create rules that allow actions, require approval or block them.
OpenAI also says background “proactive research” is restricted to read-only tools when the user is not actively working with the Dot. Actions that can affect accounts or share information can be checked through its auto-review system, and some sensitive actions, such as changing a password, remain with the user.
The company still warns that Dots can make mistakes and says consequential work should be reviewed.
That last point is not a small detail. The pitch behind these products depends on trust. People will only hand over real work if they believe the agent can act without creating a second job for them.
Meta got there first with Muse
OpenAI’s timely launch of Dots makes it impossible to separate from Meta’s Muse.
Meta introduced Muse on September 8 as a personal AI agent designed to complete tasks across the web and connected services. Muse has its own persistent virtual machine and browser, can book appointments, fill out forms, handle customer service tasks and keep working in the background. Meta also says Muse can connect to email, calendars, Instagram and other apps.
Meta has since continued expanding Muse AI. The company announced support for bringing Muse to its AI glasses, giving users a way to interact with the agent throughout the day without reaching for a phone.
OpenAI now has a product with a strikingly similar basic promise: give an agent a goal, let it work through the task and return when your attention is needed.
Then there is Grok Bot
xAI entered the same fight in August with Grok Bot, which it described as a team of “always-on agents.” Grok Bot gets its own computer, can sign into tools and apps, work across inboxes and other services, and complete jobs from start to finish before returning for approval when necessary.
xAI later expanded Grok Bot access to more SuperGrok and Cursor plans.
So Dots are not arriving in an empty market. The three products now share a strikingly similar idea: stop asking AI to help you do the work and start asking AI to do the work for you.
OpenAI’s 4,000-app bet could be the real story
The most aggressive number in OpenAI’s announcement may be the least flashy one.
OpenAI says its plugin ecosystem gives Dots access to more than 4,000 apps. That means the company is trying to make the agent useful across the software people already depend on, rather than building a closed collection of tasks inside ChatGPT.
OpenAI also has another layer planned for companies. Specialist Dots will have their own identities, credentials and defined responsibilities inside organizations. OpenAI says early internal work has included procurement, invoice processing, email marketing, customer support and commercial contracting.
The company is starting enterprise pilots and is working with Microsoft to connect specialist Dots with Agent 365 governance and security controls.
That could turn Dots into something larger than a personal assistant.
A person may start with one Dot. A company could eventually have several specialized agents handling specific responsibilities.
One Dot now, teams of Dots later
At launch, OpenAI is giving eligible users a first Dot and says it plans to support multiple Dots in the future.
Dots are rolling out to Pro and Business Premium users in eligible markets, with Enterprise users able to access a beta when their workspace administrator enables it.
OpenAI says the first Dot is included with those plans at no extra cost. Conversations with a Dot do not count against normal ChatGPT usage limits, though work handed off into Codex or ChatGPT Work can use those products’ limits.
That setup points toward the model OpenAI appears to want: one persistent agent for each person today, then a small team of agents working together tomorrow.
ChatGPT taught millions of people to ask AI for answers. Dots are OpenAI’s attempt to teach them to hand over the work itself.
That is why this launch feels bigger than another new assistant. OpenAI is no longer just competing to have the smartest thing you can talk to. It is competing to become the thing you can leave working after you walk away.
){kind=link}