- OpenAI Dots tutorial: Learn to deploy an always-on AI agent that keeps working after you log off
- Core concept: Each dot runs on its own cloud computer with browser access, powered by GPT-6 Astra
- Key skill: Set permission rules (allow, require approval, or block) before assigning projects
- Best practice: Start with one project, review the activity view, then scale up
- Availability: Rolling out to Pro and Business Premium plans, with the first dot included
What Is OpenAI Dots? Overview and Core Concepts
OpenAI Dots is a new category of always-on AI assistant introduced alongside ChatGPT. Unlike a standard chat session that stops the moment you close the tab, a dot is a persistent agent: you assign it a project, and it continues working in the background on its own cloud computer, complete with its own browser. You can walk away, lock your laptop, or switch it off entirely, and the dot keeps making progress.
Video Highlights:
- A dot can run multiple projects at once without a separate chat for each
- Dots are powered by GPT-6 Astra with dedicated cloud compute
- Dots can work together, including teammate-to-teammate collaboration
- Access to 4,000+ plugins extends what a dot can do
- A dot can contact you to ask questions or request decisions
The headline idea is autonomy without babysitting. Traditional AI agents require constant monitoring and repeated corrections. Dots invert that model: the agent works asynchronously, shares progress updates, and only interrupts you when it genuinely needs a decision.
Think of a dot less like a chatbot and more like a remote assistant with their own laptop. They work while you sleep, message you when something matters, and remember your preferences over time.
| Concept | What It Means | Why It Matters |
|---|---|---|
| Always-on assistant | Works in background after you leave | No need to monitor or re-prompt |
| Own cloud computer | Runs on OpenAI infrastructure, not your device | Works even when laptop is off |
| Plugin access | 4,000+ plugins available | Extends real-world usefulness |
| Preference learning | Learns goals and standards from feedback | Results improve over time |
| Two-way contact | Dot can message or call you | Async collaboration, not one-way commands |
Step-by-Step Setup: Activating Your First Dot
Getting started with Dots is designed to be simple, in keeping with OpenAI's ease-of-use philosophy. Follow this sequence to go from zero to a working assistant.
Confirm Plan Eligibility
Check whether your account has received the Dots rollout. Access begins with Pro and Business Premium plans. Enterprise, Education, and Healthcare users can join the beta if their workspace administrator enables it.
Locate the Dots Feature
Once rolled out, Dots appears inside the ChatGPT interface. If you do not see it yet on an eligible plan, wait for the staged rollout rather than third-party workarounds.
Configure Permissions First
Before assigning work, set access rules for apps and data: allow, require approval, or block. This is your security baseline and should be established before any real project begins.
Assign Your First Project
Give the dot a single, well-defined task with clear success criteria. A good first project is something measurable, such as preparing an invoice, tracking a schedule, or drafting recurring reports.
Review and Give Feedback
Open the activity view to see exactly what the dot did. Approve, correct, or redirect. Every piece of feedback trains the dot on your preferences for future work.
Currently, each eligible account gets one dot with an allowance for deeper work. Multiple dots working together is an announced future direction, not a feature available today. Plan your single dot's workload accordingly.
Permissions, Security, and Control Settings
Security is a major focus of the Dots design, and understanding the control system is essential before delegating anything sensitive.
Allow Rules
- Pre-approved apps the dot can use freely
- Best for low-risk, repetitive tasks
- Reduces approval interruptions
Require Approval
- Gate sensitive actions behind your sign-off
- Ideal for payments, emails, external sharing
- Dot pauses and asks before proceeding
Block Rules
- Hard restrictions on apps or actions
- Prevents unauthorized access attempts
- Review activity view for blocked events
Two additional safety layers deserve attention. First, the activity view gives full transparency into what your dot is doing at any moment, including attempts to access apps you did not authorize. Second, OpenAI operates a review system that can flag actions which might compromise your account or expose sensitive information, forcing an approval check before the action completes.
For small and medium businesses without dedicated IT departments, the activity view and automated review system provide oversight that would otherwise require dedicated staff. Check the activity log at least weekly during your first month.
| Control Layer | Function | Recommended Use |
|---|---|---|
| App permission rules | Defines what dot can access | Set before first project |
| Activity view | Live log of dot actions | Regular monitoring |
| Approval workflow | Human sign-off on flagged steps | Sensitive or financial actions |
| Automated review system | Flags risky or sensitive actions | Always-on safety net |
Practical Use Cases: What a Dot Can Do For You
Dots are positioned to help in both professional and personal contexts. OpenAI's own demonstrations highlight tasks that blend scheduling awareness, document handling, and proactive follow-through.
| Use Case | How the Dot Handles It | Best Permission Setting |
|---|---|---|
| Invoice follow-up | Notices a forgotten invoice, prepares it, sends after your approval | Require approval |
| Multi-project tracking | Manages several projects simultaneously without separate chats | Allow with activity review |
| Document drafting | Accesses local files to draft and refine recurring outputs | Allow for trusted folders |
| Team coordination | Your dot collaborates with a teammate's dot toward shared goals | Require approval for external actions |
| Personal errands | Plugins handle bookings, reminders, and research tasks | Varies by sensitivity |
Begin with a low-stakes, easily verified task such as a weekly summary or research brief. Once the dot consistently meets your standard, gradually hand over higher-stakes work. Preference learning improves results over time, but early supervision builds the feedback baseline.
A notable pattern from the demos: communication flows both ways. You can message or call your dot inside ChatGPT, and your dot can reach out to you, share progress, ask a question, or request a decision. Treat it like asynchronous collaboration rather than command-and-control.
Dots vs. Competing AI Agents
Dots arrive in a competitive landscape. The most direct comparison points available in most regions are Grok's bots (xAI) and Meta's Muse (US-only availability during initial rollout).
| Feature | OpenAI Dots | Grok Bots | Meta Muse |
|---|---|---|---|
| Always-on background work | Yes, core feature | Partial | Partial |
| Own cloud computer + browser | Yes (GPT-6 Astra) | Varies | Varies |
| Plugin ecosystem | 4,000+ plugins | Limited | Meta ecosystem |
| Two-way messaging/calls | Yes, in ChatGPT | Limited | Yes |
| Multi-agent teamwork | Demoed, future for users | Limited | Limited |
| Regional availability | Rolling out globally to eligible plans | Broader | US-focused initially |
Feature-for-feature comparisons will shift quickly as each platform iterates. Focus on your actual workflow needs: if background delegation with strong permission controls is the priority, Dots' current design directly targets that use case.
Getting Started Checklist
Your First Week With Dots:
- Confirm Dots rollout on your Pro or Business Premium account
- Configure allow / approval / block rules for all connected apps
- Assign one small, measurable first project
- Review the activity view daily during week one
- Give explicit feedback on every delivered result
- Gradually expand to multi-project delegation
Once your dot reliably completes low-stakes tasks and you trust the permission setup, you have built the foundation for true hands-off delegation. Keep feedback loops tight and let preference learning do the rest.
FAQ
Q: Does my computer need to stay on for an OpenAI dot to keep working?
No. Each dot runs on its own cloud computer with its own browser on OpenAI's infrastructure. Your dot continues working even when your laptop is asleep, locked, or powered off. You can check progress whenever you like.
Q: Who gets access to OpenAI Dots?
Dots are rolling out first to ChatGPT Pro and Business Premium users. Enterprise, Education, and Healthcare users can try the beta if their workspace administrator enables it. The first dot is included with eligible plans alongside an allowance for deeper work.
Q: Can I run multiple dots at the same time?
Not yet. Currently each account can have one dot. OpenAI has said multiple dots working together is a future direction, and demos have shown teammate dots collaborating toward a shared goal, but that is not an available feature today.
Q: How do I stop my dot from doing something I do not want?
Use the permission system to set rules per app: allow, require approval, or block. The activity view shows you everything your dot does in real time, and OpenAI's automated review system can flag sensitive or account-risky actions so they require your approval first.