OpenAI launches Dots, always-on agents with their own computers

OpenAI introduced Dots at DevDay on September 29, 2026: persistent agents, powered by GPT-6 Astra, that run on their own cloud computer and keep working on a goal after you close the chat.

Instead of answering one prompt at a time, you give a dot an objective, connect the apps it needs, and set how much it can do without asking. It works in the background, then returns results for you to review. OpenAI says dots learn from your feedback over time.

Watch the launch video

What a dot actually does

OpenAI describes dots as "remarkably capable, always-on agents built to handle everything." Each one has its own cloud computer and browser, and OpenAI says it can connect to more than 4,000 apps through plugins. With permission, a dot can also reach your laptop.

The examples OpenAI and early coverage give are everyday knowledge work: monitoring customer feedback, investigating bugs, updating documentation, revising proposals, converting content between formats, processing invoices, booking tables, and flagging financial oversights. When a dot is idle, OpenAI says it can keep doing background research using read-only tools.

You can reach a dot in ChatGPT on desktop, web, and mobile, and in Slack and Microsoft Teams. Texting is listed as coming soon. TechCrunch notes the design is deliberately friendly, a bubbly set of floating dots, and quotes OpenAI's pitch: start with a primary dot, name it, make it your own, and over time expect "teams of Dots working together on your behalf."

Who gets it, and what it costs

Your first dot is included with Pro or Business Premium at no extra cost, with extended usage limits for the first month. Conversations with a dot do not count toward usage limits. OpenAI says you will later be able to add more dots or buy more speed and monthly work. It has not published that pricing.

The model underneath matters for cost. Dots run on GPT-6 Astra, OpenAI's top model. The company launched GPT-6.1 Sol the same day as a cheaper near-Astra option for developers.

Controls and safety

An agent that acts on its own raises the obvious question of what it is allowed to do. OpenAI lists several guardrails:

• Custom Rules let you allow specific actions, require approval for others, or block them.
• Password changes always need your approval.
• Auto-review checks potentially sensitive actions against your instructions and safety requirements.
• Monitoring can pause or stop a dot if it detects a safety concern, and you can inspect what a dot did on its cloud computer at any time.
• Isolation keeps the dot on a separate cloud computer, with defenses against malicious instructions it might meet while browsing.

For Business and Enterprise workspaces, content is excluded from model training by default. Personal-plan users control that in their settings.

These are OpenAI's descriptions of its own safeguards. Nothing in the launch materials tells you how often they fail, so treat them as design intent until independent testing appears.

Dots for organizations

OpenAI is also pitching dedicated "specialist" dots that have their own identity and credentials inside a company's systems. It says early pilots cover procurement, invoicing, email marketing, support, and contracting, and that it is working with Microsoft on integration with Agent 365 governance controls.

What to do with it

If you are on Pro or Business Premium, start with one narrow, low-risk recurring job, such as a weekly summary of customer feedback, and keep approvals on for anything that sends messages or touches money. Check the activity log after the first few runs. Teams on Enterprise should decide who owns a dot's credentials and what it may access before an admin enables the beta. The broader shift is the one already visible in ChatGPT Work: the product is moving from a chat window to a worker that holds a task for you.