OpenAI Dots: What These New Always-On AI Agents Actually Do for You
For years, AI assistants have been stuck in a loop: you ask, they answer, and then they forget you exist until the next prompt. At its DevDay 2026 keynote on September 29, OpenAI tried to break that loop with Dots — always-on AI agents that keep chasing your goals in the background, whether you’re at your desk or asleep.
If you’ve ever wished an assistant would just go do the thing instead of waiting for you to spell out every step, this is aimed squarely at you. Here’s what OpenAI Dots actually are, what they can do today, and whether they’re worth your attention.
What Are OpenAI Dots?
Dots are personal, always-on AI agents powered by a new model called GPT-6 Astra. The pitch, in OpenAI’s words, is “remarkably capable, always-on agents built to handle everything.” Instead of living inside a single chat window, a Dot runs continuously in the background, pursuing goals you define with minimal oversight.
The branding is intentionally soft — bubbly, cartoonish floating dot avatars meant to make an autonomous agent feel less like a robot and more like a helpful sidekick. You start with your primary Dot, give it a name, and make it your own. OpenAI’s longer-term vision is “teams of Dots working together on your behalf,” including specialist Dots tuned for specific business roles.
The key mental shift: a chatbot answers questions, a Dot does work. It’s the difference between “tell me how to file this” and “file this.”
How Dots Actually Get Work Done
Here’s the part that makes Dots different from a normal chat assistant: each one gets its own cloud computer and browser. That means a Dot can open web pages, use connected tools, and click through tasks the same way you would — without borrowing your machine (though you can grant access to your personal computer when a job needs it).
Because it runs continuously, a Dot can also handle recurring work, follow up on tasks over time, and even delegate pieces of a job to sub-agents. When it’s not actively working on something you assigned, it can quietly do read-only research across your connected tools so it’s ready when you need it.
You stay in the loop the whole time: OpenAI lets you watch the agent’s cloud computer while it works, drop in feedback, or even hop on a voice call to steer it mid-task.
Real Examples: What You’d Assign a Dot
Abstract “autonomous agent” talk is easy to tune out, so here are the concrete use cases OpenAI is leading with:
- For developers: point a Dot at your customer feedback channel, and it can investigate bugs mentioned in Slack and implement fixes.
- For builders: hand it a finished design, and it can spin up the app when the design is ready.
- For operations: let it generate invoices for your approval without being prompted each time.
- For researchers: have it rerun analyses and investigate unexpected experimental results on its own.
Notice the pattern: these are the repetitive, “someone should really automate this” tasks that eat your week. A Dot is built to own the whole loop — notice, act, follow up — rather than handing each step back to you.
Where You Can Talk to a Dot
Dots aren’t locked to one app. You can assign work and check in through ChatGPT’s desktop and browser apps, Slack, and Microsoft Teams, with text-message support on the way. Setup happens through ChatGPT first, and after that you can interact from mobile too.
Under the hood, Dots connect to over 4,000 apps through OpenAI’s plugin ecosystem — the same connection framework ChatGPT already uses. That’s what lets a Dot pull information from your connected services and act on a task instead of waiting for you to feed it every detail.
Staying in Control: Permissions and Safety
Handing an AI the keys to your tools is exactly where most people get nervous, and OpenAI knows it. Dots inherit ChatGPT’s existing permission framework, with a few extra guardrails layered on top.
You can set custom rules that decide whether a Dot acts on its own or has to pause and ask before proceeding. Consequential actions can require confirmation, and there’s activity monitoring plus automated reviews of the riskier moves. OpenAI also used DevDay to introduce Private Intelligence, including a preview of private inference using confidential computing — a nod to the obvious question of what an always-on agent does with your data.
The practical takeaway: start any new Dot with tight rules that require approval, watch how it behaves, and only loosen the leash once it’s earned your trust. Treat it like onboarding a new teammate, not flipping a magic switch.
Can You Use Dots Yet? Availability and Pricing
Dots launched at DevDay for Pro and Business Premium subscribers in eligible markets, with Enterprise users able to turn them on with admin approval. One Dot is included with those plans, and you can add more agents later. Creating one requires the ChatGPT desktop or browser app.
Notably, OpenAI also used the same event to announce a new $500-per-month Pro tier alongside GPT-6.1 Sol and expanded Codex developer tools — a sign of just how much compute these always-on agents can burn through. If you’re on a free or Plus plan, you’ll be watching from the sidelines for now.
Should You Care About Dots?
Dots are the clearest sign yet that AI is graduating from “answer my question” to “do my work.” An assistant that runs 24/7, uses real tools, and follows up on its own is genuinely useful — and genuinely worth being thoughtful about. If you have access, the smart move is to start small: pick one repetitive, low-stakes task, wrap it in strict approval rules, and see how much time it actually gives back.
The tools that amplify you the most aren’t the ones that do the most magic — they’re the ones you learn to direct well. A Dot is only as good as the goals and guardrails you hand it. Get those right, and you’re not replacing yourself; you’re cloning your best hours.
Sources & further reading:
- OpenAI launches Dots, its bubbly agentic avatar (TechCrunch)
- OpenAI launches Dots, new ‘always-on agents’ you can assign tasks to (9to5Google)
- OpenAI DevDay 2026: Dots, GPT-6.1 Sol & more (Business Standard)
Related Reading
- OpenAI GPT-5.5: The AI Super App That Changes Everything About How You Work
- AI Agents for Everyday Users: How to Set Up Your First Automated Assistant in 2026
- AI Agents Are Finally Going Mainstream: Your Personal Productivity Revolution Starts Now