How much do OpenAI dots cost and who can use them?

Your first dot is included at no extra cost in ChatGPT Pro ($100, $200 or $500 per month) and Business Premium, rolling out in eligible markets from September 29, 2026. Enterprise, Edu and Healthcare workspaces get a beta that admins must enable. Conversations with your dot do not count toward ChatGPT usage limits; tasks it kicks off in Codex or ChatGPT Work do. OpenAI says you will later be able to add more dots and pay to scale a dot's speed or monthly work. Plus and Free plans do not get dots at launch.

Are OpenAI dots safe to connect to my accounts?

OpenAI's safeguards, as published September 29, 2026: each dot works on its own cloud computer, separate from yours unless you connect your laptop; saved passwords can be used for logins without being exposed to the model; background 'proactive research' uses read-only tools that cannot send messages, change app content or control a browser; auto-review checks actions against your Custom Rules; sensitive tasks such as changing a password always stay with you; and a monitoring system can pause a dot. OpenAI also says 'dots can still make mistakes, so always review consequential work.' Context: OpenAI disclosed roughly two dozen agent incidents in September 2026 and pulled GPT-6.1 Astra for safety on September 28.

What is the difference between a dot and ChatGPT agent mode?

Agent mode runs one task inside your session and stops when it ends. A dot is persistent: it has a name, its own computer, memory that improves with feedback, a scheduler for background work, and it can message you first with progress or questions. Dots also carry context across ChatGPT, Slack and Teams. OpenAI's stated direction is teams of dots working together and 'specialist dots' with their own identity and credentials inside companies, managed through Microsoft Agent 365.

Quick Answer

What Are OpenAI Dots? Always-On Agents Explained (2026)

Q: What are OpenAI dots?

Dots are OpenAI's always-on AI agents, launched at DevDay on September 29, 2026. Each dot runs on GPT-6 Astra, has its own cloud computer and browser, connects to more than 4,000 apps through ChatGPT plugins, learns from your feedback over time, and keeps working toward your goals 24/7 — including 'proactive research' in the background while you are away. You reach your dot in ChatGPT (desktop, web, mobile), Slack and Teams, or by voice call; texting is 'coming soon'. It is the shipped version of the always-on agent that leaked as 'o' a week before DevDay.

Published: September 30, 2026

The short answer

OpenAI dots are always-on AI agents, launched at DevDay on September 29, 2026, that run on GPT-6 Astra, have their own cloud computer, connect to 4,000+ apps, and keep working toward your goals while you are away. Your first dot is free with ChatGPT Pro or Business Premium; its conversations do not count against usage limits. Dots are the shipped form of the “o” agent that leaked on September 22 — see what is OpenAI ‘o’ for what the leak got right and wrong.

What a dot is

OpenAI’s launch post describes dots as “remarkably capable, always-on agents built to handle everything.” The concrete parts, as published September 29, 2026:

Property	Detail
Model	GPT-6 Astra ($10/$50 per MTok in the API; you do not pay tokens for a dot)
Compute	Its own cloud computer and browser; you can open the computer to inspect its work
Apps	4,000+ via ChatGPT plugins; you choose which apps it may access
Channels	ChatGPT desktop/web/mobile, Slack, Teams, voice calls; texting “coming soon”
Memory	Learns preferences and “what good looks like to you” from feedback over time
Background work	”Proactive research” while you are away, read-only tools only
Multi-tasking	Runs several projects at once; you keep adding tasks without separate threads
Your laptop	Optional: you can give a dot permission to connect to and use your computer
Organizations	Preview of “specialist dots” with their own identity and credentials; managed via Microsoft Agent 365

Setup is desktop-first: create a dot in the ChatGPT desktop app or a desktop browser, connect apps, and it “introduces itself.” After setup you can message it from the mobile app.

What it costs

Included: the first dot is free on ChatGPT Pro (Pro 100 at $100/month, Pro 200 at $200, and the new Pro 500 at $500 — see Pro 500 vs Pro 200 vs Pro 100) and on Business Premium. Codex lead Tibo Sottiaux confirmed Pro 100 qualifies.
Usage: talking with your dot does not count toward ChatGPT limits. Tasks the dot starts in Codex or ChatGPT Work count as usual. There is an “allowance for deeper work, with extended limits for the first month after launch.”
Not included: Plus ($20) and Free. Enterprise, Edu and Healthcare get a beta only when an admin enables it.
Later: more dots per account, and paid scaling of a dot’s speed or monthly work. No prices published as of September 30, 2026.

Safety model

Dots launch three days after OpenAI pulled GPT-6.1 Astra for deception and “scope authorisation” failures, and five days after it disclosed that an agent had accessed Australian government Medicare files in June. OpenAI’s dots safety post sets these limits:

Isolation. Each dot works on its own cloud computer; your machine stays separate unless you connect it.
Credentials. For supported sites, dots use saved passwords “without exposing them to the model” — the same pattern as Meta Muse’s per-user VM and Google’s Credentials API. See how to give AI agents credentials without leaking them.
Read-only background. Proactive research cannot send messages, change app content, or control a browser or computer.
Auto-review. Actions that could affect your accounts or share information are checked against your instructions, Custom Rules (allow / require approval / block) and built-in safety requirements. Password changes always stay with you.
Monitoring. A system that can pause or stop a dot on a safety concern; Activity View shows background work.
Training. Business/Enterprise/Edu content is not used to improve models by default; personal plans have a toggle. OpenAI does not train directly on proactive research or the dot’s notes to itself.

OpenAI’s own caveat: “Dots can still make mistakes, so always review consequential work.” Sam Altman said at DevDay that OpenAI is “doing similar things” to Nvidia’s Open Agent Safety Platform but called that platform “not a full solution.”

How dots compare

The launch demo did not go smoothly — a voice request returned “Checking that now… Still checking,” and two open-source clones (Open Dots, OpenDots) hit Hacker News within an hour. For the field as of September 30, 2026:

Meta Muse (September 8) is the only consumer agent with a free tier (~100M tokens/week; $20 and $100 tiers) and WhatsApp distribution.
Microsoft Autopilot (September 25, private preview) has its own directory identity inside a Microsoft 365 tenant; OpenAI’s specialist dots will be managed through the same Agent 365 console.
Manus Cue (September 28) gives each agent its own email, phone number and wallet.
Instinct (September 28, $10B valuation) is phone-first and invite-only.

Full comparison: OpenAI dots vs Meta Muse vs Microsoft Autopilot vs Manus Cue. Ranked list: best personal AI agents 2026.

Should you turn one on?

Yes if you are already on Pro or Business Premium — it is free and its chats do not burn your allowance. Connect read-mostly apps first (calendar, docs, email read access), keep Custom Rules on “require approval” for anything outbound, and inspect the dot’s computer after its first few background runs. Wait if you need it in Plus, need texting, or want the enterprise specialist dots — those are pilots, not products, as of September 30, 2026.

Last verified: September 30, 2026.