OpenAI Dots Explained: What Are These Always-On AI Agents?
For years, using an AI assistant has mostly meant opening a chat window, asking a question, and waiting for an answer. OpenAI’s new Dots are built around a different idea: give the AI an ongoing responsibility and let it continue working after you leave the conversation.
OpenAI introduced Dots on September 29, 2026, at DevDay as “always-on” AI agents designed to work across projects, connected apps and communication tools. They are powered by GPT-6 Astra, have their own cloud computer and browser, and can connect to more than 4,000 apps through OpenAI’s ecosystem.
That changes the relationship between a person and an AI assistant. Instead of repeatedly explaining what needs to happen next, a Dot is intended to remember the context, make progress on a project, and come back when it needs a decision or approval.
But “always-on” does not mean the AI has unlimited freedom to do anything it wants. OpenAI has built permission controls, custom rules, action reviews and safeguards into Dots, particularly for actions that could affect accounts, data or other people.
So, what exactly are OpenAI Dots, how do they work, and what makes them different from regular ChatGPT?
Quick Answer: What Are OpenAI Dots?
OpenAI Dots are persistent AI agents designed to take responsibility for ongoing tasks rather than simply answering individual prompts. They are powered by GPT-6 Astra and operate using their own cloud computer, browser and connected apps.
The “always-on” part means a Dot can continue making progress on assigned work between conversations. It can research information, monitor connected sources, work through multi-step projects and send updates when it needs your input.
Users can interact with Dots through ChatGPT and, according to OpenAI, through Slack and Microsoft Teams. Dots can also use connected applications and carry context across communication channels.
At launch, Dots are rolling out to eligible Pro and Business Premium users, while Enterprise users can access a beta when their workspace administrator enables it. Availability varies by market.
What Are OpenAI Dots?
The simplest way to understand OpenAI Dots is to think of them as ongoing AI workers rather than one-off chat sessions.
A normal ChatGPT conversation generally starts when you ask something. You provide a prompt, the model responds, and you decide what happens next.
A Dot is designed to take a goal and continue working toward it.
OpenAI describes Dots as agents that can take on ongoing responsibility, work across several projects and operate independently within boundaries set by the user.
Imagine you are running a website.
Instead of asking ChatGPT:
“What topics are trending in AI this week?”
and then manually continuing the research, you could give a Dot a broader responsibility such as monitoring customer feedback, researching relevant developments and preparing updates for you to review.
The difference is subtle at first, but important. The human gives the direction; the agent handles more of the work between those moments.
OpenAI’s own examples include developers asking Dots to watch customer feedback, identify recurring problems, build and test fixes, and prepare pull requests for review. Other examples involve updating launch materials as requirements change, revising research as new evidence arrives, and preparing content from new interview transcripts.
How Do OpenAI Dots Work?
Several technologies come together to make Dots possible.
GPT-6 Astra
Dots are powered by OpenAI’s GPT-6 Astra model. That distinction matters: Astra is the underlying AI model, while a Dot is the agentic system built around that model.
The model provides the intelligence needed to understand instructions, reason through problems and work with tools. The Dot adds persistence, connected applications, a cloud computer, permissions and an ongoing task structure.
Their Own Cloud Computer
Each Dot has its own cloud computer and browser. This gives the agent a working environment where it can carry out tasks instead of merely describing what a person should do.
Users can also open the Dot’s computer to inspect its work.
Local computer access is separate. OpenAI says access to a user’s own computer starts turned off and must be explicitly enabled. Once connected, a Dot can access files and perform work on that computer according to the permissions granted.
Connected Apps
Dots can connect to more than 4,000 apps through OpenAI’s plugin ecosystem. The exact capabilities depend on the apps a user chooses to connect and the permissions available to those integrations.
That makes connected apps a major part of the concept.
An agent that can read relevant information from your calendar, project tools, documents or communication platforms has much more context than an isolated chatbot.
Memory and Context
Dots are also designed to learn from feedback and remember context for ongoing work.
OpenAI says a Dot can learn a user’s preferences, standards and working style over time. It can also carry context between ChatGPT, Slack and Teams.
That is particularly useful for projects where the instructions are not completely written down in a single prompt.
The Basic Workflow
A simplified version looks like this:
You give a goal → the Dot gathers context → plans the work → uses its computer and connected apps → makes progress → asks for decisions when needed → returns completed work for review.
The important word is “simplified.” Real tasks can involve several loops of research, tool use, checking and revision.
What Can OpenAI Dots Actually Do?
OpenAI’s examples suggest that Dots are intended for work that normally requires several separate actions.
Research and Monitoring
A Dot can research a topic, monitor information and continue investigating as new information becomes available.
OpenAI describes a “proactive research” mode where Dots can use information from connected apps in the background. During this background work, connected-app tools are restricted to read-only access, meaning they cannot use those tools to send messages, modify app content or control a computer.
Software Development
This is one of the clearest examples from OpenAI.
A developer could have a Dot watch customer feedback, identify smaller bugs or feature requests, scope the work, build and test changes, and prepare pull requests for human review.
Dots can also create Codex cloud tasks, although those tasks use the relevant Codex allowances associated with the user’s plan.
Content Production
OpenAI describes a workflow in which a Dot receives an interview transcript, identifies useful moments for clips, prepares show notes and drafts social posts.
The idea is not simply to generate text. The Dot can maintain the context of the project and carry edits or preferences across related work.
Project Management
Dots can also be used for projects where requirements change.
For example, if a product launch changes direction, a Dot can work through what those changes mean for existing materials and prepare updated drafts.
The practical benefit is that the user does not have to restart the entire process every time something changes.
What Does “Always-On” Actually Mean?
This is probably the most misunderstood part of OpenAI Dots.
“Always-on” does not mean that a Dot is necessarily thinking continuously every second without interruption.
Instead, it describes an agent that can maintain ongoing responsibility and continue working between active conversations.
OpenAI says Dots can work toward user goals 24/7 and perform proactive research in the background.
There are also important boundaries.
When Dots perform proactive research while the user is not actively working with them, OpenAI says those tools are restricted to read-only access. This prevents that background mode from independently sending messages or changing connected app content.
So “always-on” is better understood as persistent availability and ongoing responsibility, rather than unlimited autonomous access.
OpenAI Dots vs Regular ChatGPT
|
Feature |
Regular ChatGPT |
OpenAI Dots |
|
Interaction model |
Primarily prompt-and-response conversations |
Ongoing agent relationship |
|
Ongoing tasks |
Usually requires continued interaction |
Designed to continue work between conversations |
|
Background work |
Limited depending on feature |
Proactive/background work is a core feature |
|
Connected apps |
Available through ChatGPT’s app ecosystem |
Uses connected apps as part of ongoing tasks |
|
Proactive behavior |
Generally waits for the user |
Can identify follow-up work and send updates |
|
Computer access |
Depends on the specific ChatGPT feature |
Each Dot has its own cloud computer |
|
Memory/context |
Depends on ChatGPT features and settings |
Designed around persistent project context |
|
User approvals |
Depends on the task |
Built-in rules and approval controls for sensitive actions |
|
Best use case |
Questions, conversations and individual tasks |
Delegating ongoing, multi-step responsibilities |
The distinction is not that one replaces the other.
Regular ChatGPT remains useful when you want an answer or need to work interactively. Dots are designed for situations where the work continues after the initial conversation.
OpenAI Dots vs AI Agents: What Makes Them Different?
The word “agent” has become one of the most common terms in AI.
But an agent is more than a chatbot with a fancy name.
A useful AI agent generally has several characteristics:
- it can pursue a goal rather than just answer a question;
- it can use tools;
- it can observe information from its environment;
- it can decide what action to take next;
- it can perform multiple steps;
- it can maintain context across those steps;
- and it can operate with some degree of autonomy.
Dots fit that model.
Their significance comes from putting those capabilities into an ongoing assistant that can remain responsible for a project instead of ending after one response.
OpenAI’s broader DevDay announcements reinforce this direction. The company described its new systems as giving people ways to take on ongoing responsibilities and creating shared environments where humans and agents can collaborate.
That represents a broader shift in AI: from “ask the model to do something” toward “give the agent responsibility for something.”
OpenAI Dots vs Meta Muse
OpenAI launched Dots at a time when Meta’s Muse was already attracting attention in the personal-agent market.
The two products overlap in important ways, but they are not identical.
|
Category |
OpenAI Dots |
Meta Muse |
|
Core idea |
Always-on AI agent for ongoing responsibilities |
Personalized AI agent designed for ongoing assistance |
|
AI model |
GPT-6 Astra |
Meta has not publicly disclosed an equivalent model name in the sources reviewed |
|
Computer environment |
Own cloud computer and browser |
Muse has also been expanded toward computer interaction |
|
App integration |
More than 4,000 connected apps claimed by OpenAI |
Integrations vary by Muse experience |
|
Work communication |
ChatGPT, Slack and Teams |
Meta’s consumer and business ecosystem |
|
Business focus |
Business Premium and Enterprise availability from launch |
Meta expanded Muse toward small-business use |
|
Personalization |
Learns preferences and working style |
Personalized agent and customizable experience |
|
Pricing |
Included with eligible plans at launch |
Free for most uses, with paid tiers reported by Axios |
|
Availability |
Gradual rollout in eligible markets |
Availability varies by Muse market |
Meta expanded Muse into small-business use on September 29, describing use cases such as analyzing sales and social activity, helping with cash flow and inventory, drafting customer communications and creating marketing content. Axios reported that Muse remained free for most uses, with $20 and $100 monthly tiers available.
OpenAI’s Dots, meanwhile, are tied closely to the company’s ChatGPT, Codex, connected-app and enterprise ecosystem. OpenAI says Dots can use more than 4,000 apps and communicate through ChatGPT, Slack and Teams.
The important difference is therefore less about a simple “winner” and more about the ecosystems surrounding each agent.
OpenAI Dots Price and Availability
As of September 30, 2026, Dots are in an initial rollout rather than being universally available.
OpenAI’s current Help Center documentation says Dots are rolling out to ChatGPT Pro users in eligible markets, while Business Premium users can access them across supported ChatGPT regions. Enterprise users, including Edu and Healthcare workspaces, can try the beta when their workspace administrator enables it.
For Pro users, OpenAI says Dots are currently unavailable in the European Economic Area, Switzerland and the United Kingdom. Business Premium availability is described separately as covering supported ChatGPT regions.
The first Dot is included at no additional charge with eligible Pro and Business Premium plans. OpenAI says the first month after launch includes extended limits for deeper work, while longer-term usage terms are to be provided for each plan.
There is another important detail: conversations with a Dot do not count toward normal ChatGPT usage limits, but tasks started or managed through Codex or ChatGPT Work count toward those services’ usage allowances.
OpenAI says users will eventually be able to add more Dots and scale the amount or speed of work, but pricing for additional Dots had not been announced in the launch information reviewed for this article.
Because the rollout is happening gradually, seeing no Dot option in an eligible account does not necessarily mean the account is permanently excluded. OpenAI says access may take several days to reach an account.
OpenAI Dots and Privacy: What Should Users Know?
The more responsibility an AI receives, the more important permissions become.
OpenAI has therefore made access control a central part of Dots.
Connected Apps
Users choose which apps a Dot can access and can manage those permissions through ChatGPT’s existing app controls.
This matters because connected apps can contain sensitive information.
A Dot should therefore have access only to the information needed for the work it has been assigned.
Local Computer Access
A Dot’s cloud computer is separate from the user’s own computer.
OpenAI says local computer access starts disabled. If the user explicitly connects a computer, the Dot can access files and work on that computer until the permission is revoked.
That separation is an important part of the security model.
Custom Rules
Users can create Custom Rules that specify what a Dot can do independently, what requires approval and what is blocked.
OpenAI says built-in safety requirements still apply even when custom rules are configured.
Sensitive Actions
Dots use an action-review system called auto-review to evaluate actions against user instructions, Custom Rules and safety requirements.
Some sensitive actions, such as changing a password, remain with the user.
OpenAI also says business, enterprise and education workspace content is not used to improve its models by default. Personal-plan users can control whether their Dot conversations and work are used for model improvement.
The Risks of Always-On AI Agents
The benefits of persistent agents come with an obvious trade-off: the more an AI can do without asking, the more carefully its boundaries need to be designed.
One risk is excessive permissions.
If a Dot has access to email, files, calendars and other applications, a mistake can have consequences beyond an incorrect chatbot answer.
There is also the possibility of an agent misunderstanding instructions or acting on incomplete information.
Prompt injection is another concern for agentic systems. An agent that browses websites or reads external documents can encounter instructions designed to manipulate its behavior. OpenAI says Dots include safeguards intended to defend against malicious instructions and monitor for potentially harmful behavior.
The broader industry is also dealing with real-world examples of agent failures. Reuters reported around the Dots launch that OpenAI had been examining rogue-agent activity following earlier incidents involving agents and external systems.
That context explains why approval systems matter.
An AI that simply gives you a wrong paragraph is annoying.
An AI that changes a password, sends a sensitive email or modifies important files without permission can create a much bigger problem.
For that reason, OpenAI itself says Dots can still make mistakes and advises users to review consequential work.
Why OpenAI Is Moving Toward Always-On Agents
Dots are not an isolated product idea.
The wider AI industry is moving from conversational models toward systems that can use tools, operate computers and complete multi-step tasks.
OpenAI’s DevDay 2026 announcements reflect this direction. The company described new ways for people and AI to work together, including ongoing agent responsibilities and shared environments where humans and agents can collaborate.
The economic reason is fairly straightforward.
A chatbot helps when someone is actively using it.
An agent can potentially remain useful between interactions.
That means the AI becomes part of a workflow rather than simply being another application people open when they need information.
For businesses, that could mean agents handling recurring research, documentation, coding tasks, customer-support preparation or internal processes.
For individuals, it could mean having an assistant that remembers what matters and keeps projects moving.
The important change is not simply that models are becoming smarter. It is that software is increasingly being designed to act on the user’s behalf.
What Could OpenAI Dots Change?
The longer-term implications are still uncertain, but several possibilities are already visible.
Personal Productivity
A persistent agent could handle recurring research, reminders, planning and administrative tasks without requiring a fresh prompt every time.
Small Businesses
A business owner could potentially delegate research, customer communication drafts, marketing preparation, document work or operational analysis to an agent.
Software Development
Dots’ integration with coding tools makes software development one of the clearest areas where persistent agents could change workflows.
Instead of asking an AI to write one function, a developer could give it responsibility for investigating bugs, preparing fixes and returning work for review.
Research
For researchers, the ability to monitor new information and update analyses could reduce some of the repetitive work surrounding a project.
Workplace Collaboration
OpenAI’s ChatGPT Space is designed as a shared environment where humans and AI can collaborate. Dots can participate in these workflows, making agents more like members of a digital work environment rather than isolated chatbots.
The Agent Ecosystem
Perhaps the most interesting possibility is the emergence of teams of specialized agents.
OpenAI says it envisions users eventually being able to work with teams of Dots, while its specialist-dot initiative is already being tested for defined organizational responsibilities.
That is still a developing direction, not a finished reality.
But it suggests where the technology could go next: one general assistant coordinating with several specialized agents, each responsible for a particular type of work.
Frequently Asked Questions
What are OpenAI Dots?
OpenAI Dots are persistent, always-on AI agents designed to take responsibility for ongoing tasks. They are powered by GPT-6 Astra and have their own cloud computer, browser and access to connected apps.
Are OpenAI Dots always running?
Dots are designed to work toward assigned goals continuously and can perform background or proactive research. However, “always-on” does not mean unlimited autonomous access. Background research uses restricted read-only tools, and sensitive actions can require approval.
What model powers OpenAI Dots?
OpenAI Dots are powered by GPT-6 Astra. Astra is the underlying model; Dots are the agent system built around it.
Can OpenAI Dots use my apps?
Yes. Dots can use apps that the user chooses to connect, subject to the permissions and capabilities of those integrations. OpenAI says its ecosystem provides connections to more than 4,000 apps.
Can Dots access my computer?
A Dot has its own cloud computer. Access to a user’s local computer is optional and starts disabled. Users must explicitly connect and grant access.
How much does OpenAI Dots cost?
The first Dot is included at no extra charge with eligible Pro and Business Premium plans. Enterprise users can access the beta when their administrator enables it. OpenAI has said additional Dots will eventually be available, but pricing for extra Dots was not specified in the launch announcement.
Are OpenAI Dots available everywhere?
No. As of September 30, 2026, rollout is gradual. OpenAI’s Help Center says Pro availability excludes the European Economic Area, Switzerland and the UK, while Business Premium and Enterprise availability follows separate supported-region and workspace rules.
How are Dots different from ChatGPT?
Regular ChatGPT is primarily designed around interactive conversations and individual tasks. Dots are designed to maintain ongoing responsibility, work in the background and continue projects between conversations.
Are OpenAI Dots safe?
OpenAI has built Dots with permissions, Custom Rules, action review and safety monitoring. However, OpenAI also acknowledges that Dots can make mistakes and recommends reviewing consequential work.
Can Dots work without me?
They can continue working on assigned projects and perform certain background tasks without an active conversation. However, they can stop and ask for human input when a decision or approval is required.
Can I control what a Dot is allowed to do?
Yes. Users can choose connected apps, configure permissions and create Custom Rules that allow, block or require approval for specific actions.
Are OpenAI Dots available for free users?
As of September 30, 2026, OpenAI’s current documentation does not list Free users among the initial Dots rollout. The current rollout covers eligible Pro and Business Premium users, with Enterprise beta access when enabled by an administrator.
Final Thoughts
OpenAI Dots represent a meaningful change in how an AI assistant is expected to behave.
The traditional chatbot model is simple: you ask, it answers, and you decide what happens next.
Dots try to move one step beyond that. You give the agent a goal, provide the necessary context and let it continue working. It can research, use connected apps, operate a cloud computer, remember the project and return when it needs your judgment.
That is what makes the “always-on” idea interesting.
At the same time, autonomy creates a new responsibility for both the company building the agent and the person using it. Permissions, approval rules and human review become much more important when AI can actually take actions instead of simply generating text.
As of September 30, 2026, Dots are still at the beginning of their rollout, with availability limited by plan and region. But the direction is clear: AI assistants are moving from systems we constantly prompt toward systems that can take responsibility for work over time.
The real question now is less about whether an AI can answer a question, and more about which parts of your ongoing work you would actually trust an AI agent to handle.