Dots by OpenAI

About Dots by OpenAI

Dots are always-on agents inside ChatGPT, each running on a dedicated cloud computer with its own browser. Built on GPT‑6 Astra, a dot connects to over 4,000 apps through plugins and works toward your goals 24/7. You can message or call your dot in ChatGPT on desktop, web, and mobile, or reach it through Slack and Teams.

Review

Dots represents a shift from on-demand AI assistance to persistent, proactive agents that keep working between conversations. Each dot operates its own virtual machine, so it can browse the web, use connected apps, and prepare work while you're away. The agent learns from your feedback and surfaces finished work for your review, with Custom Rules and an Activity View governing what it can do on its own.

Key Features

• Dedicated cloud computer and browser. Every dot runs in its own isolated environment, allowing it to perform multi-step tasks across websites and apps without sharing resources with other agents.
• 24/7 proactive work. Dots don't wait for prompts - they spot tasks that need attention and bring completed work to you. One early tester reported their dot noticing a forgotten invoice, preparing it, and sending it after approval.
• Cross-platform access. You can interact with your dot through ChatGPT on desktop, web, and mobile, plus voice calls and messaging inside Slack and Teams.
• Custom Rules and Activity View. These controls let you set boundaries on what the dot does autonomously and review a log of its actions.
• Feedback-based learning. The dot adjusts its behavior over time based on the feedback you give it, refining how it handles your preferences and workflows.

Pricing and Value

Dots is rolling out now to ChatGPT Pro and Business Premium subscribers. OpenAI hasn't published separate pricing for Dots, which suggests it's included in those existing plans rather than sold as an add-on. Whether lower-tier plans will gain access later hasn't been confirmed.

Pros

• Runs on its own virtual machine, which means it can execute long-running tasks without tying up your local device or browser session.
• Learns from explicit feedback, so repeated corrections shape how the agent handles similar work in the future.
• Approval gates exist for certain actions - the invoice example shows the dot prepared the work and waited for human sign-off before sending.
• Works inside tools where teams already communicate (Slack, Teams), reducing the need to switch contexts.
• Activity View gives a transparent record of what the dot did on its own, which helps with trust and auditing.

Cons

• Access is limited to Pro and Business Premium plans at launch; users on lower-tier plans don't have it yet.
• The scope of approval gates isn't documented publicly - it's unclear which action types (payments, outbound emails, data deletion) require human confirmation and which the dot might execute based on its own confidence level.
• This tool isn't well suited for people who prefer fully manual control over every task. If you're uncomfortable delegating work to an agent that acts between conversations, the always-on nature may create more anxiety than it relieves.

Dots fits professionals who already delegate recurring digital tasks and want an agent that keeps moving projects forward without constant prompting. The dedicated VM and app plugin system make it practical for multi-step workflows across different services. Teams that rely heavily on Slack or Teams for coordination may find the messaging integration useful, though anyone on a lower-tier ChatGPT plan will have to wait for broader availability.