Dots
PaidAlways-on agents that keep working while you're away.
Published 1 October 2026
Scores
Popularity3/5
Launched at DevDay with wide press coverage and ChatGPT's distribution behind it, yet access is limited to the top paid plans and rollout is still reaching accounts, so real usage is small next to ChatGPT itself or Meta's free Muse.
Learning Curve1/5
Set up from inside ChatGPT with no infrastructure, keys, or model choice, and used by chatting or calling it like a colleague in apps people already have open.
Flexibility3/5
Thousands of plugin connectors plus Slack and Teams presence cover a wide range of work, but behaviour is shaped only through feedback and approval rules, with no code-level extension or choice of model.
Performance3/5
Runs on GPT-6 Astra with a full cloud computer per agent, but it has only just begun rolling out and there are no independent task evaluations or sustained user reports to judge reliability on long unattended work.
Portability1/5
Exists only inside ChatGPT on OpenAI's infrastructure and models, with memories that cannot be exported and no self-hosted or API path.
About Dots
Dots are OpenAI's always-on personal agents, announced at DevDay on September 29, 2026 and rolling out inside ChatGPT. A dot is a named agent that learns a user's preferences from feedback, runs multi-step projects on its own, and keeps working toward the user's goals when nobody is chatting with it, looking for useful background work between conversations. They run on OpenAI's GPT-6 Astra model.
Each dot has its own cloud computer that the user can open at any time to watch what it is doing. From there it can browse, use apps, research, draft documents, and write software, and it reaches more than 4,000 services through ChatGPT's plugin ecosystem. A dot can also connect to the user's own machine for local tasks. Users can message or call a dot in ChatGPT on web, desktop, and mobile, work with it inside Codex, or talk to it in Slack and Microsoft Teams, and it carries the same context across every channel. Voice calls work from launch; text messaging and dots calling the user are planned for later releases.
Actions that touch a user's accounts or share information have to clear an auto-review check first, saved passwords are used without being exposed to the model, apps can be restricted to read-only, a monitoring system can pause or stop an agent over a safety concern, and other actions follow allow, block, or ask-first rules. Memory spans apps, but individual memories cannot yet be viewed or edited, which is the main privacy gap early coverage points out.
Dots are not sold separately. One dot is included with every ChatGPT Pro tier and with Business Premium seats; Enterprise, Edu, and Healthcare workspaces get a beta once an admin enables it. Pro users in the EEA, Switzerland, and the UK are excluded for now, while Business Premium covers all supported ChatGPT regions. Specialist dots with their own credentials and teams of dots are planned, at prices OpenAI has not disclosed. It is OpenAI's answer to Meta's Muse: a managed personal agent with no self-hosting or model choice, opposite open-source agents such as OpenClaw and Hermes Agent.
Key Features
- Persistent named agents that learn preferences from feedback
- A dedicated cloud computer per dot that the user can open and watch
- Background work toward goals between conversations
- Reachable by chat or voice call in ChatGPT, Codex, Slack, and Microsoft Teams
- Connects to more than 4,000 apps through ChatGPT plugins
- Auto-review, read-only app limits, and hidden saved passwords for account actions
- Powered by GPT-6 Astra
Pros
- Lives where many teams already work, with the same agent answering in ChatGPT, Slack, and Teams
- Inherits ChatGPT's plugin ecosystem, far more connectors than rival personal agents offer at launch
- Watchable cloud computer makes it easy to see what the agent is doing and step in
- No extra charge for the first dot on eligible plans
Cons
- Needs ChatGPT Pro or Business Premium, so the cheapest entry is $100 a month
- Pro users in the EEA, Switzerland, and the UK are excluded at launch
- Stored memories cannot yet be viewed, edited, or deleted individually
- Launched with a staggered rollout and no independent track record yet
Dots Pricing
Paid- · One dot included at no extra charge
- · A $200 Pro tier raises usage limits
- · Not available to Pro users in the EEA, Switzerland, or the UK
- · Per seat, or $100 per seat billed annually
- · One dot included per seat
- · Available in all supported ChatGPT regions
- · One dot included
- · 25 times the ChatGPT Plus allowance
- · Only Pro plan with the Ultrafast speed mode
- · Dots beta once a workspace admin enables it
- · Also covers Edu and Healthcare plans
Tools Related to Dots
Integrates with Dots(2)
A dot can be worked with inside Codex, carrying the same memory and context it has in ChatGPT, Slack, and Microsoft Teams.
Dots run on OpenAI's GPT-6 Astra model and live inside ChatGPT, with no option to use another vendor's model.
Alternatives to Dots(7)
Meta Muse and OpenAI's Dots are the two big-vendor personal agents, each giving the user an agent with its own cloud computer. Muse has a free plan but is limited to the US and Canada; Dots come only with ChatGPT Pro or Business Premium and plug into Codex, Slack, and Teams.
Dots and Grok Bot are both always-on agents bundled into a larger subscription rather than sold separately: Dots live in ChatGPT and suit an individual's ongoing projects, while Grok Bot is aimed at company messaging, approvals, and workflow automation.
Dots and Manus are both managed agents that work in their own cloud computer. Dots are persistent, keep working between conversations, and come bundled with ChatGPT Pro; Manus is a standalone product with a free tier, focused on turning one delegated task into a finished deliverable.
Both can be messaged from Slack and Microsoft Teams. A dot is one person's agent that follows them across ChatGPT and Codex; Viktor is a shared coworker for the whole workspace, billed from a common credit pool instead of per seat.
Dots are managed by OpenAI and run only on its models inside ChatGPT; Hermes Agent is open source and self-hosted, so the agent, its memory, and the model choice stay on your own server.
Dots are OpenAI's managed agents with no model choice and a ChatGPT Pro requirement; OpenClaw is free and open source, runs on hardware you control, and works with any model provider.
Vendor
OpenAI
Website →Tags
Details
- Maintained
- Yes
- Agent type
- Autonomous app
- Open source
- No
- Hosting
- Cloud