OpenAI dots: review and open-source alternative
A plain description of OpenAI's always-on agent inside ChatGPT, the facts worth knowing before you rely on it, an honest review, and a side-by-side comparison with ControlClaw.
Updated November 7, 2026. OpenAI dots details come from their public pages. Spotted something out of date? Tell us.
What is OpenAI dots?
Dots is OpenAI's always-on agent inside ChatGPT. You create one dot, give it a name and a goal, and it keeps working between conversations. It has its own cloud computer and browser. It uses the apps you have already connected to ChatGPT, and it can message you in ChatGPT, Slack or Teams. When you are not talking to it, it reads your connected apps in the background and looks for things to do.
It is closed source and fully hosted. The dot, its memory and the models it uses all live on OpenAI's servers. The models are OpenAI's own (GPT-6 Astra and GPT-6.1 Terra); you cannot plug in another provider. You do not pick a server, and only Enterprise workspaces can choose a data region.
Dots is OpenAI's entry into the category OpenClaw created in the open: the persistent personal agent. The comparison below is really a comparison of two ways to get one: a closed service inside ChatGPT, or open-source software run for you with guardrails. ControlClaw agents can run on OpenAI's models too, so the choice is not about whose model answers. It is about who runs the agent and who sets the rules.
OpenAI dots at a glance.
What OpenAI has published, and what it has not.
Dots comes with a ChatGPT plan rather than being sold on its own. ChatGPT Pro ($100, $200 or $500 a month) includes the first dot. So does a Business Premium seat ($100 a user a month billed yearly, $125 billed monthly). Enterprise, Edu and Healthcare workspaces got dots on November 4, 2026. Billing there starts on November 18, at flat rates or by usage, and OpenAI has not published those prices. Each person gets one dot. OpenAI says more dots per person and "specialist dots" for companies are coming, but has not said when or at what price. Pro users in Europe (the EEA and Switzerland) and the UK cannot use dots yet. Business Premium and Enterprise users can, in supported regions.
Memory is shared with ChatGPT's memory, and the dot adds its own notes from your apps. You cannot view or delete single memories. To clear them, you delete the dot. There is no export format. Safety is built in. Default rules say which actions need approval. Custom Rules let you allow, block or require approval for an action. Auto-review checks risky actions before they run, and a monitoring system can pause the dot. Some things, such as changing a password or moving money, the dot hands back to you. On Enterprise, admins can turn dots on or off per role, pause or delete any dot, and read audit logs. OpenAI has not published admin controls for dots on Business Premium.
- Type
- Hosted, closed-source always-on agent inside ChatGPT
- Agents per plan
- One personal dot per person; more planned
- Where it runs
- OpenAI's cloud; a data region only on Enterprise
- Models
- OpenAI's GPT-6 family only
- Memory
- Shared with ChatGPT; single memories cannot be viewed or deleted; no export
- Setup
- Minutes in the ChatGPT app
- Safety controls
- Default rules, Custom Rules, Auto-review, safety monitoring
- Team admin
- Enterprise only: roles, pause, delete, audit logs
- Pricing
- First dot included in Pro ($100 to $500 a month) and Business Premium; Enterprise rates not published
- Open source
- No
OpenAI dots review: capable, well guarded, and OpenAI's to run
We run the competing product, so read this with that in mind. We have tried to be fair; where dots is better, we say so.
Dots is the most complete hosted agent we have seen from a big lab. It gets its own computer and browser. It uses the thousands of app connections ChatGPT already has, and you can reach it from ChatGPT, Slack or Teams with the same context everywhere. The safety work is real. Custom Rules have four levels, from "take action without asking" to "hand off to you". Auto-review checks an email's recipient before it goes out. A login form keeps passwords away from the model. For one person on a Pro plan who already lives in ChatGPT, it is a strong product, and the price is a plan you may already pay for.
The trade is control. You cannot see the code, choose a model from another provider, or pick where the dot runs. A data region exists only on Enterprise. OpenAI holds the memory and mixes it with ChatGPT's. You cannot read or delete a single memory, and there is no export, so leaving means starting over. The rules are good, but they are OpenAI's rules, judged by OpenAI's models. You can tell a dot never to send email. You cannot give it a list of domains it may write to. Business Premium has no published admin console for dots. On personal plans, OpenAI can use the dot's conversations and work to train its models unless you turn that off.
There is also a regional gap. Pro users in Europe and the UK cannot create a dot at all. And the question that hangs over every hosted agent hangs over this one. If OpenAI changes or retires dots, the dot and its memory go with it. An open-source agent on a server you can export from does not have that problem.
What dots does well
- Own cloud computer and browser; works while you are away
- Thousands of app connections already set up in ChatGPT
- Clear action rules: allow, pre-approve, ask, or hand off
- Auto-review checks risky actions such as sending email
- Passwords go to the browser, not the model
Where it falls short
- Closed source; OpenAI models only
- No data region unless you are on Enterprise; no Pro access in Europe or the UK
- Memory mixed with ChatGPT's; no view, edit or export
- One dot per person; Business admin controls not published
- Personal-plan data may train OpenAI's models unless you opt out
A very good personal agent if you already pay for ChatGPT Pro or Business Premium and are happy for OpenAI to hold the agent, its memory and the rules. Not a fit when you need to pick the model or the region, write your own allow-lists, or keep the option to leave with your agent intact.
Both work overnight. Only one leaves with you.
A persistent agent that works while you sleep
Both keep running without a laptop open, remember context and act across your tools.
Open and portable versus closed and hosted
ControlClaw runs OpenClaw you can read, fork and move, on a model from any provider, OpenAI included. A dot runs OpenAI's code on OpenAI's models, on servers you cannot pick.
Teams with rules versus one person in ChatGPT
A dot is one assistant per ChatGPT seat. ControlClaw is a fleet with allow-lists, approvals and a shared console.
Feature comparison: ControlClaw vs OpenAI dots
| Feature | ControlClaw | OpenAI dots |
|---|---|---|
| Ownership | ||
| Agent code | YesOpen-source OpenClaw, on GitHub | NoClosed source |
| Where it runs | Hetzner or AWS in EU, US, UK, Singapore, or your own cloud | OpenAI's cloud; a data region on Enterprise only |
| Models | Any provider, or your own LLM subscription | OpenAI's GPT-6 family |
| Memory and data | YesYours; export any time, dedicated server per agent | PartialHeld by OpenAI; shared with ChatGPT memory |
| Leaving | YesExport the agent and run OpenClaw anywhere | NoNo export of the dot's memory |
| Safety | ||
| Secrets | YesHeld by the Firewall, never on the agent | PartialHeld by OpenAI; a login form keeps passwords from the model |
| Prompt-injection screening | YesInbound content inspected, rules you set | PartialModel training and Auto-review; not rules you write |
| Approvals | YesOne queue, full context, audit trail | YesCustom Rules, Auto-review, Activity View |
| Email and calendar allow-lists | YesBy domain, address or group | PartialRules such as "never send email"; no list by domain published |
| Spend and tool limits | YesPer agent, enforced outside the agent | PartialPlan allowance; website allow and block lists on Enterprise |
| Capabilities | ||
| Always on | YesDedicated server per agent | YesOwn cloud computer |
| Own mailbox with AgentMail or your Gmail | Your personal email account; no own address at launch | |
| Meetings and calls | YesJoins Meet, Zoom, Teams and phone calls | PartialYou can call it in ChatGPT; it cannot call you; joining meetings not published |
| Multiple agents | YesFleet, templates, shared Brain | PartialOne dot per person; more planned |
| Browser and passwords | YesVaulted; the agent never sees them | YesOwn browser; saved passwords hidden from the model |
| Commercial | ||
| Price | From $19/mo including a Small server | Included in Pro from $100/mo or a Business Premium seat from $100/mo |
| Data location | EU, US, UK or Singapore, your pick | PartialEnterprise only; nine storage regions, EU or US inference |
| Trial | Fourteen days, no charge, card required | NoNo free trial published |
"Not published" means we could not find it documented. If OpenAI publishes it, we will update the row.
OpenAI dots or ControlClaw: match the job.
Pick OpenAI dots
You already pay for ChatGPT and want an assistant inside it.
- You are on Pro or Business Premium outside Europe and the UK
- You are happy with OpenAI's models and OpenAI's rules
- One dot per person is enough
- Memory can stay with OpenAI
- You want ChatGPT's app connections without setup
Pick ControlClaw
You need to know where it runs and what it may do.
- Open-source code you can audit and fork
- EU or other data location on request
- Allow-lists and approvals before it emails or pays
- A team console for more than one agent
- The option to leave with your agent intact
OpenAI dots vs ControlClaw pricing.
No per-message pricing on ControlClaw. Agents, seats, the firewall and starter AI usage are in the plan; bring your own LLM subscription if you have one.
OpenAI dots vs ControlClaw: common questions.
Is there an open-source alternative to OpenAI dots?
Yes. OpenClaw is the open-source always-on agent, and ControlClaw runs it for you with an AI Firewall, approvals and a team console, on servers in the region you choose.
Can a ControlClaw agent use OpenAI's models?
Yes. Agents use any provider, or your own LLM subscription, OpenAI included. The difference is who runs the agent and who sets the rules, not whose model answers.
Can I move my dot's memory to ControlClaw?
Not directly. Dots has no export. Copy what you need from the dot's conversations and files and paste it into the agent's memory.
Can I use dots in Europe or the UK?
Not on a personal Pro plan. Business Premium and Enterprise workspaces can, in supported regions, and Enterprise can keep the dot's data in Europe. ControlClaw runs in the EU on every plan.
Why does a firewall matter if dots already has Auto-review?
Models follow instructions, including hidden ones in a web page or email. Dots checks an action after the model has read that content. The Firewall screens the content before the agent sees it, and holds the credentials the agent would need to do damage.
Sources (11)
- https://openai.com/index/introducing-dots/
- https://help.openai.com/en/articles/20001530-getting-started-with-your-dot
- https://help.openai.com/en/articles/20001529-dots-privacy-security-and-safety-faqs
- https://help.openai.com/en/articles/20001554-managing-dots-in-chatgpt-workspaces
- https://help.openai.com/en/articles/9903489-data-residency-and-inference-residency-for-chatgpt
- https://help.openai.com/en/articles/9793128-about-chatgpt-pro-tiers
- https://help.openai.com/en/articles/8792828-chatgpt-business-overview
- https://chatgpt.com/pricing
- https://www.engadget.com/2281998
- https://www.nbcnews.com/tech/tech-news/openai-launches-dots-ai-agents-safety-questions-rcna600338
- https://thenextweb.com/news/openai-dots-always-on-ai-agents-cloud-computers-devday