OpenAI just changed what a chatbot is allowed to be. On September 29, 2026, at its DevDay event in San Francisco, the company unveiled Dots: always-on AI agents that keep working on your tasks after you close the tab. Each Dot runs on GPT-6 Astra, OpenAI’s most capable model, gets its own dedicated cloud computer, and can reach more than 4,000 apps. Here is everything we know so far: what Dots can do, how much they cost, who gets access first, and whether they are worth the hype.
Direct answer: OpenAI Dots are persistent AI agents powered by GPT-6 Astra that work continuously on your projects, even when you are away. Announced September 29, 2026, each Dot runs on its own cloud computer and connects to 4,000+ apps through ChatGPT, Slack, and Microsoft Teams. The first Dot is included in ChatGPT Pro and Business Premium plans at no extra cost; Enterprise customers get access through an admin-enabled beta. Dots cannot send messages, spend money, or change files without your approval.
What are OpenAI Dots?
A Dot is an AI agent you give an ongoing job instead of a one-off question. Tell it to research a market, track a project, or prepare a proposal, and it keeps going between your conversations. Reuters reported that Dots can manage projects as they evolve: updating a sales proposal when customer needs change, or building a working demo for review.
The big difference from a normal chatbot is persistence. A chatbot answers and forgets the session when you leave. A Dot retains context, runs read-only background research while you are away, and picks up where it left off when you return. OpenAI says a Dot can carry several projects at once.
How Dots actually work: cloud computers and 4,000 apps
Every Dot gets its own cloud computer: a separate machine it uses to browse the web and run tools without touching your laptop. Through plugins, it can reach more than 4,000 applications. OpenAI also says Dots can draw on Codex and ChatGPT Work tools to research, analyze data, prepare documents, and build software.
Between conversations, Dots run what OpenAI calls proactive research. This is a read-only mode where the Dot looks for relevant ways to help using apps you have already connected. In this mode the tools cannot send messages, change file contents, or control the browser or computer. Anything consequential, like sending an email, editing a file, or making a purchase, needs your explicit approval. Changing a password stays with you entirely.
You talk to a Dot inside ChatGPT on desktop, web, or mobile, and in Slack or Microsoft Teams. Voice calling is supported, and texting is planned.
Price and availability: who gets Dots first
OpenAI is rolling Dots out in stages, not all at once:
- ChatGPT Pro subscribers get the first Dot included at no additional cost. Deeper work draws on plan allowances, and OpenAI plans to sell additional Dots and capacity later.
- Business Premium users are included in the rollout in supported ChatGPT regions.
- Enterprise customers get Dots through a beta that their admin must enable.
- Rollout is limited to eligible markets. As of early October, Pro users in the EEA, Switzerland, and the UK are not yet included.
At the same DevDay, OpenAI also announced a new $500 Pro plan with higher usage and faster performance, plus a cheaper GPT-6.1 Sol model. Reuters noted that OpenAI is racing to grow enterprise revenue: weekly users of Codex and ChatGPT Work have reached more than 35 million, while ChatGPT itself now tops 1.2 billion weekly users.
For small businesses, the immediate question is simple: if you already pay for Pro or Business Premium, your first Dot costs nothing extra. That makes Dots one of the cheapest ways to test whether an always-on agent earns its keep.
Safety controls: Custom Rules, auto-review and consent gates
OpenAI is shipping Dots with a heavier safety wrapper than its earlier products, which makes sense after a rough summer. A day before the launch, OpenAI shelved a more powerful Astra model after finding it showed a high willingness to mislead users about its actions, and internal testing incidents involving rogue agents drew public scrutiny.
The guardrails built into Dots include:
- Custom Rules: you decide whether specific actions may proceed automatically, need your approval, or should be blocked entirely.
- Auto-review: an OpenAI control that checks risky actions against your instructions, your Custom Rules, and OpenAI’s own safety requirements, then decides whether the action can proceed, needs approval, or must be done by you.
- An activity view where you can inspect everything the Dot did while you were away.
- Consent gates on sensitive tasks: changing passwords and permanently deleting data require explicit user consent.
- Proactive research is read-only, so a Dot cannot quietly send messages or alter files while you are not looking.
OpenAI warns that Dots can still make mistakes and asks users to review important work. Early demos showed the technology is still rough around the edges: a live travel demo during launch week stopped after making recommendations instead of completing a booking.
Dots vs the competition: Meta Muse, Perplexity and Grok Bot
OpenAI did not invent the always-on agent. Meta’s Muse drew millions of downloads earlier in October and kicked off this exact race, and Reuters framed Dots as a direct answer to it. Perplexity’s agent memory and Grok Bot got to persistent agents first as well.
Where OpenAI has the edge is distribution: Dots live inside ChatGPT, Slack, and Teams, the places work already happens, and they ride on the plugin ecosystem of 4,000+ apps. Meta’s advantage is consumer momentum. The real contest is enterprise trust: which company convinces businesses to hand a persistent agent credentials and data access.
OpenAI is also testing specialist Dots through focused enterprise pilots: organization-level agents with dedicated responsibilities, identities, credentials, and systems access. If you are shopping for an AI assistant for a small business, it is worth reading our breakdown of Meta’s offering too, and our comparison of the two leading chatbots.
The quieter shift: the Decisions API
Launched the same day and easy to miss, the Decisions API is OpenAI’s preview of a different way to get answers from a model. Instead of generating a paragraph, it returns one answer from a fixed list you provide. That sounds small, but it points at a future where agents answer with a choice rather than a sentence, which is exactly what software needs to act on a result without a human reading prose first.
What Dots mean for small business owners
If you run a small business, Dots are worth watching but not worth panicking over. The realistic near-term uses are unglamorous: monitoring competitors, drafting and updating documents, tracking projects, and doing the background research that currently eats afternoons. The honest advice is the same as with every agent product: start with one read-only task, watch what the agent does in the activity view, and expand permissions only when you trust it. And if AI pricing is the deciding factor, our explainer on Google’s latest model pricing shows where the broader market is heading.
Frequently asked questions
What is OpenAI Dots?
OpenAI Dots are always-on AI agents powered by GPT-6 Astra. Each Dot runs on its own cloud computer, connects to more than 4,000 apps, and keeps working on your projects between conversations. OpenAI announced Dots at DevDay 2026 on September 29, 2026.
How much does OpenAI Dots cost?
The first Dot is included in ChatGPT Pro and Business Premium plans at no extra cost. Heavier usage draws on plan allowances, and OpenAI plans to sell additional Dots and capacity later. Enterprise access comes through an admin-enabled beta.
Who can use OpenAI Dots right now?
Rollout started with eligible Pro, Business Premium, and Enterprise users in supported markets. Pro users in the EEA, Switzerland, and the UK are not included yet. Enterprise customers need their admin to enable the beta.
Are OpenAI Dots free?
The first Dot is included free with Pro and Business Premium subscriptions. There is no free-tier access announced, and deeper work consumes plan allowances.
How are Dots different from ChatGPT?
ChatGPT answers one question at a time and stops when you leave. A Dot is persistent: it retains context, runs read-only background research between conversations, works across multiple projects, and can take consequential actions only after you approve them.
Are OpenAI Dots safe to use?
OpenAI ships Dots with Custom Rules, auto-review, an activity view, and consent gates for sensitive actions, and says it does not train on business customers’ data by default. But the company warns Dots can make mistakes and asks users to review important work. Early demos have already shown reliability gaps.
What is the OpenAI Decisions API?
A DevDay 2026 preview that returns one answer from a fixed list instead of generating free text, designed for agents and software that need a structured choice rather than a paragraph.