A working definition of OpenAI Dots
A useful definition is short enough to argue with. Here it is. In community discussion, OpenAI Dots is an agent from OpenAI, aimed at people who already use ChatGPT, that can take a goal and work on it with tools. The tools people name most often are a virtual machine or other computer, Slack, and email. The account people mention next is ChatGPT, sometimes specifically Pro, because heavy computer use does not fit the mental model of a free chat box. That is the whole definition this site is willing to print. It is not a feature matrix, a ship date, or a price.
The definition has two layers, and mixing them is how a guide becomes fan fiction. The first layer is what OpenAI has published. Workspace agents in ChatGPT can run in the cloud, can be shared on a team, can run on a schedule, and can be placed in Slack. The Agents API documents an OpenAI-hosted sandbox: a Linux workspace with a shell, files, and a network policy the application sets. Those products are real, scoped, and written down by OpenAI. The second layer is the community word Dots, used when people want a single name for a personal or team agent that feels like both of those ideas plus a mailbox. This page keeps the layers in separate paragraphs on purpose.
Dots vs Muse
Muse is Meta’s personal agent. Public descriptions put one agent next to one person: mail, calendar, travel, shopping, and goals that last longer than a single chat. Muse talks about a secure virtual machine for that person, with credentials kept away from the main model, and about approvals when an action leaves the machine. The product has its own apps. It is not a mode inside ChatGPT, and it is not sold by OpenAI.
Dots vs Muse is therefore a comparison of jobs, not of versions. Muse, as Meta describes it, is a steward for a life. The Dots conversation is a steward for work that already lives in ChatGPT: a thread, a file, a channel, a follow-up. Both pictures include a computer. Both pictures include messages. They do not include the same default audience. A household choosing Muse is choosing Meta’s apps and Meta’s connectors. A person asking about Dots is usually asking whether their OpenAI account can grow an agent without adopting a second assistant with its own memory.
The honest difference in openness is small. Both are closed products. You do not host the main Muse runtime yourself, and this guide does not pretend you host Dots yourself. What you can compare today is the published story. Muse publishes a personal-agent story. OpenAI publishes workspace agents and developer sandboxes. The community then asks whether Dots is the name that will eventually sit over a personal OpenAI agent with a similar computer. That question is open. The table below marks it as community orientation so it cannot be copied into a slide as a spec.
If you are choosing a tool this week, choose from what you can sign in to. Read Muse’s own site for Muse. Read OpenAI Dots features on this site only as a map of the jobs people want — VM, Slack, email — and then read OpenAI’s workspace-agent pages for the Slack path that is actually documented. Do not let a comparison table outrank a vendor’s current plan list. This site does not print Muse’s prices. Those belong to Meta and they move.
Dots vs Grok Bot
Grok Bot, in public product talk from SpaceXAI / xAI, is a roster. The pitch is teammates rather than a single steward: persistent agents, each with a computer, able to keep a workflow going for a team. People who have used it describe bots that learn a task by watching the normal interface of a piece of software, and a coordinator that hands work between them. That is a different shape from one ChatGPT thread with one machine attached.
Dots vs Grok Bot is a comparison of cardinality. The Dots conversation almost always has one agent identity tied to an OpenAI account. Grok Bot’s conversation has many identities, on purpose, so a team can separate roles. Hosting rhymes — a computer per agent, files, a browser — and then the organizational model diverges. If your problem is “I want five roles and a handoff,” Grok Bot is the product whose public story matches the sentence. If your problem is “I want the assistant I already pay OpenAI for to operate a computer and answer in Slack,” you are in the Dots conversation, and you should still verify which of those clauses OpenAI has published.
Openness does not rescue either side. Grok Bot is not an open-source runtime you clone and run. Dots, as discussed, is not a public repository of OpenAI’s agent. Comparing them on “who is more open” usually hides the real choice: which account holds the credentials, which admin can turn the agent off, and which log you can read after a bad action. Those questions are on the capabilities page as checklist rows, because they are preferences you can write down before a vendor answers them.
Prices are out of scope for this table. Grok Bot’s commercial terms are whatever xAI publishes this week. ChatGPT’s plans are whatever OpenAI publishes. A Dots subscription, if one exists as a separate line item, has not been established on this site. See OpenAI Dots pricing for the guess cards and for the ChatGPT figures this guide is willing to repeat.
Dots vs Muse vs Grok Bot: comparison table
The control above the grid highlights one column or restores all three. Dimmed columns stay in the page so you can still read them. Identity, hosting, openness, audience, and integrations are the rows because those are the questions that change a buying decision. Benchmarks are absent on purpose. A community guide that invents a score is worse than one that refuses.
What OpenAI has published
Workspace agents are an evolution of shared custom agents in ChatGPT. OpenAI’s introduction says they run in the cloud, keep working when you are away, and can be used together in ChatGPT or Slack. The help center describes creating an agent, connecting Slack with a handle and a channel, running on a schedule, and the requirement that Slack use shared authentication rather than a stray personal login. Availability in that preview is tied to business and education plans, with admin controls. That is a specific product with a specific gate. It is the right citation when someone says “OpenAI already has a Slack agent.”
Hosted sandboxes are a developer surface. The hosted sandbox documentation describes a Linux workspace OpenAI provisions, with Python, Node.js, and command-line tools, a working directory, and network access that can be open, closed, or limited to a host list. Files can persist for the life of a session and selected outputs can be published as artifacts. That is the right citation when someone says “an OpenAI agent can have a VM.” It is an API environment your application asks for. It is not, by itself, a button in the ChatGPT sidebar named Dots.
You can hold both citations and still not know what a consumer client will call the combination. That gap is why this page exists. It is also why the page refuses to fill the gap with a screenshot tour. For the names people use while the gap is open, read OpenAI Aeon, orbit, and O.
What is still a community question about Dots
Several practical questions stay in the community layer until a public doc answers them. Is email in the same client as Slack, or is email a connector some teams will add and others will not? Does a personal ChatGPT plan include any computer time, or is the computer reserved for a workspace admin? Can a thread hand work to the machine without a separate developer session? What is the audit log a teammate can export? Who can approve a send, a purchase, or a delete?
None of those questions is answered by choosing a row in the table. They are answered by a product page, a help article, or a contract. Until then, write the questions into the checklist as things you care about, and treat any confident grid you see elsewhere as someone’s hope. The Chinese edition of this definition is at 什么是 OpenAI Dots.
How to use the answer tomorrow
- If a vendor demo shows a computer, ask whose account owns the machine and how you revoke it.
- If a demo shows Slack, ask whether the path matches OpenAI’s published workspace-agent Slack, or whether it is a different client.
- If a demo shows a price, do not paste it into a budget until you can open an OpenAI page that states it. Start at pricing.
- If a demo uses the word Aeon, orbit, or O, read the names guide before you repeat it as a brand.
Frequently asked questions
What is OpenAI Dots?
OpenAI Dots is the community name for an OpenAI agent that people describe as having a computer, a ChatGPT thread, and a way to show up in Slack or email. It is a name for a conversation, not a published product specification.
What is the difference between Dots and Muse?
Muse is Meta’s personal agent for one person’s life admin, with its own apps. The Dots conversation is about an agent for work that already lives in ChatGPT. Both pictures include a computer; they do not share a default audience or vendor.
What is the difference between Dots and Grok Bot?
Grok Bot, from SpaceXAI / xAI, is described as a roster of teammate agents, each with its own computer. The Dots conversation almost always means one agent identity tied to an OpenAI account.
Has OpenAI published anything like Dots?
Yes, in pieces. OpenAI has published workspace agents in ChatGPT, which can run in the cloud, be shared, run on schedules, and work in Slack, and hosted sandboxes in the Agents API, which give developers a Linux workspace. Neither is titled Dots.