# What your Dot can access and do

> OpenAI Dots in ChatGPT: current Pro and Business Premium prices, regional access, setup steps, permissions, and how Codex and Work tasks use your allowance.

- Published: 2026-09-30
- Last reviewed: 2026-09-30
- Data as of: 2026-09-30
- Article slug: openai-dots-guide
- Author: [Glevd](https://x.com/glevd)
- Reading time: 9 minutes
- Topics: openai, dots, chatgpt, ai-agents, pricing, permissions
- Data policy: Dated analysis; static benchmark claims are retained.
- Canonical URL: https://benchlm.ai/blog/posts/openai-dots-guide

OpenAI Dots can keep working between chats with a cloud computer and connected apps. OpenAI began rolling them out on September 29, 2026. [Launch announcement](https://openai.com/index/introducing-dots/).

The first OpenAI Dot is included with eligible ChatGPT Pro or Business Premium. Dot conversations are included. Tasks delegated to Codex or ChatGPT Work use their usual allowances.

We checked the documentation on September 30. Our recommendation is to start with one responsibility you can review, then check access, permissions, and usage before adding more work.

## Your plan decides access before usage

Dots access depends on the plan, region, and account rollout. At launch, Pro users in the European Economic Area, Switzerland, and the UK are excluded. Business Premium users can qualify across supported ChatGPT regions. Enterprise, including Edu and Healthcare, has an administrator-enabled beta that starts switched off. The [getting-started article](https://help.openai.com/en/articles/20001530-getting-started-with-your-dot) gives these conditions and says access can take several days. Dots are also [unavailable to users under 18](https://help.openai.com/en/articles/20001529-dots-privacy-security-and-safety-faqs).

The published USD subscription prices below were checked September 30. They buy a plan or seat, with the first Dot included in eligible Pro and Business Premium access.

| Plan or seat | Published subscription price | What decides whether it fits |
| --- | --- | --- |
| Free, Go, or Plus | No qualifying Dots plan listed at launch | Buying usage credits does not establish Dots eligibility |
| Pro 100 | $100 per month | Lowest-priced personal Pro tier; check region and included usage before upgrading |
| Pro 200 | $200 per month | More included usage than Pro 100; existing and new subscriptions can have different allowances |
| Pro 500 | $500 per month | Highest included usage among Pro tiers; Dots inclusion alone does not require this tier |
| Business Premium | $125 per seat/month on monthly billing, or $100 per seat/month billed annually | Managed workspace and Premium seat eligibility; account and workspace controls apply |
| Enterprise, Edu, or Healthcare | Contract pricing | Administrator must enable the Dots beta; verify the workspace's access and usage terms |

Prices and tier distinctions come from [ChatGPT Pro tiers](https://help.openai.com/en/articles/9793128-about-chatgpt-pro-tiers) and the [Business overview](https://help.openai.com/en/articles/8792828-chatgpt-business-overview). Regional prices and currency can vary.

A new Business workspace needs at least two paid seats, which can mix Standard and Premium. One Premium plus one Standard totals $150 per month on monthly billing, or $120 per month billed annually: $125 + $25, or $100 + $20. Those are seat-price sums, before any applicable taxes or additional usage. A $100 Premium seat is therefore part of a workspace purchase with an annual commitment.

For an eligible individual who wants to try Dots, we'd start with the lowest Pro tier that covers their other work. A team needing workspace administration has a different purchasing decision. Use the account's usage figures to judge whether its allowance can cover your month of tasks.

If you're comparing Dots with another agent product, the [personal AI agents hub](/ai-agents) puts access, permission controls, and billing rules side by side.

Existing Pro 200 subscribers should check their notice. Eligible subscribers can retain their previous included allowance through October 29, 2026. New subscriptions without that eligibility receive the revised allowance. The price remains $200. Avoid budgeting from an old allowance multiplier without checking which terms apply to the account.

## Included conversations do not make delegated tasks unlimited

OpenAI includes an allowance for deeper Dot work and describes extended limits during the first month after launch. Its current launch pages do not give a fixed number of tasks, hours, or credits for that allowance. Additional Dots and ways to scale their output are future plans, with no price specified there.

One usage distinction is explicit: [talking to a Dot does not count toward ChatGPT usage limits](https://openai.com/index/introducing-dots/). Asking it to start or manage tasks in Codex or ChatGPT Work counts toward those products' limits as usual. Having a Dot start the task does not turn that task into an unmetered conversation.

Work and Codex share the plan's included usage pool. Task size, model choice, reasoning settings, and the work performed affect consumption. Their [usage guide](https://help.openai.com/en/articles/20001516-managing-usage-with-gpt-6-astra-in-work-and-codex) says the allowance cannot be converted into a fixed task count. Check Settings → Usage for remaining allowance and resets. Where both five-hour and weekly limits apply, both must have room.

For example, a Dot conversation about a proposed bug fix and a Codex task to implement it have different accounting. Check the delegated task's usage before turning that request into a recurring responsibility. One successful task tells you what that task used. It doesn't establish the cost of every future fix.

Eligible personal accounts can use [ChatGPT usage credits](https://help.openai.com/en/articles/12642688-using-credits-for-flexible-usage-in-chatgpt-personal-plans) after their included allowance. Availability depends on the feature, plan, and account. Review any automatic reload and its monthly cap before leaving recurring work running. These credits are separate from API credits.

Our [model pricing table](/llm-pricing) compares API token rates. It can't calculate a Dot subscription's task allowance. Keep that distinction when comparing a ChatGPT plan with an agent you operate through a model API.

## Desktop setup keeps messaging and app access separate

These steps follow [OpenAI's setup documentation](https://learn.chatgpt.com/docs/dots/getting-started). We haven't tested the onboarding in an account.

1. Sign in to the account and workspace you intend to use, then check the access conditions above.
2. Open Dots in the ChatGPT desktop app or a desktop browser and follow the introduction. You can personalize the name and appearance.
3. Review connected apps and add the sources the first task needs. App connections can be skipped during setup and added later.
4. Decide whether the task needs your computer. A local connection is optional. Start with the cloud computer if the work doesn't need local files or tools.
5. Give the Dot a responsibility, its source material, and the decisions to bring back. Open its profile and Activity to inspect delegated work and requests for input.

Local computer access [starts turned off](https://help.openai.com/en/articles/20001530-getting-started-with-your-dot). Connect it from the desktop app on that computer if you need it. Keep the device online with ChatGPT open while the Dot uses it. That requirement applies to local work. The Dot's separate cloud computer can keep working when your device is off.

To add a messaging channel, open the Dot's profile and select Add. [Slack and Microsoft Teams](https://learn.chatgpt.com/docs/dots/channels) are contact methods for the same Dot. Connecting one does not automatically connect an inbox, grant other plugin access, or connect your computer.

Be explicit about monitoring. Adding a Dot to a Slack channel doesn't create a schedule. Specify the channel, what to watch, when to check, and which changes should prompt an update. Ask it to confirm the saved recurring task, then inspect that task's instructions and destination.

Texting has narrower availability: the [Help Center](https://help.openai.com/en/articles/20001530-getting-started-with-your-dot) describes a limited US Pro beta where offered, excluding Business and Enterprise workspaces. Treat it as an account-specific option rather than a required setup step.

## Permissions distinguish research from actions

Dots use GPT-6 Astra and have their own cloud computer and browser, as documented in [Meet dots](https://learn.chatgpt.com/docs/dots). Access to your accounts still depends on the connections you permit.

Plugin permissions are [shared across Dots, ChatGPT, Work, and Codex](https://help.openai.com/en/articles/20001529-dots-privacy-security-and-safety-faqs). Review existing connections during setup, not just the new ones. A previously connected account can already be available within its granted permissions.

OpenAI calls the Dot's unsolicited background research “proactive research.” Its tools are read-only: they can gather permitted information and save private notes, but cannot directly message others, change app content, or control a browser or desktop. Work you explicitly authorized can continue in the background under the usual action rules. Any follow-up action from research needs those rules and checks too. [OpenAI's safeguards explanation](https://openai.com/index/how-we-build-safety-security-and-privacy-into-dots/) separates these cases.

Custom Rules set boundaries for supported actions. On desktop, the [controls guide](https://learn.chatgpt.com/docs/dots/controls) directs you to Settings → Personalization → Custom rules under Permissions. For a first customer-facing task, requiring approval before sending messages gives you a point to review the recipient and content. A request to prepare a draft should say whether sending is authorized.

Dots' built-in confirmations and handoffs still apply. Purchases with saved merchant cards require approval. Permanent deletion and new security-sensitive access require confirmation. Changing a password or transferring money between financial accounts must be handed back to you. Custom Rules cannot remove these mandatory protections. Auto-review checks relevant actions against instructions and safety requirements, and Dots can still make mistakes.

For an eligible Enterprise workspace, administrators enable the beta in Workspace Settings → Permissions & roles → Permissions → Workspace default → Workspace capabilities. Local computer access and permission to add or edit Custom Rules have separate controls and start off for Enterprise. Default action rules still apply. Enabling Dots does not grant access to every app; app approval and each connected service's authorization still apply. The [Enterprise workspace guide](https://help.openai.com/en/articles/20001554-manage-dots-in-chatgpt-workspaces) covers these controls. Those Enterprise defaults should not be assumed for a personal Pro account or a Business workspace.

Pausing also has a specific scope. The controls guide says Pause stops the Dot's current main task. To stop delegated work, open that task in Activity. To end a recurring run, disable or delete its saved schedule in Scheduled. Check all three when ending a responsibility. Closing a conversation or call doesn't necessarily end the work.

Before connecting private material, review the data settings. Personal plans use the “Improve the model for everyone” setting for eligible Dot conversations and work. Business, Enterprise, and Edu workspace content isn't used for training by default. Disconnecting a plugin stops new access through that connection but doesn't erase information already in the Dot's context. The [privacy FAQ](https://help.openai.com/en/articles/20001529-dots-privacy-security-and-safety-faqs) explains those controls.

## Start with a job you can check

Our first-trial recommendation is a responsibility with a visible result and a clear approval boundary. For example, ask a Dot to review one project's updates and prepare a list of changed deadlines with links to the source messages. Decide whether it should merely report the changes or also edit a shared document. That choice changes the permissions to review.

A setup prompt could read:

> Review the connected planning channel each weekday at 9 a.m. Eastern for two weeks. Prepare a private list of changed deadlines with source links. Ask me before editing shared files or messaging anyone. Notify me when a deadline moves or an owner is missing. Confirm the saved schedule and its destination.

This is a suggested prompt, not a tested workflow. Change the source, timing, and permission boundary to match the work you actually want done.

| Check before the first recurring run | What to inspect |
| --- | --- |
| Access | Correct account, qualifying plan or seat, region, and workspace enablement |
| Billing | Remaining Work/Codex allowance, any purchased-credit use, and automatic reload settings |
| Sources | Connected accounts and permissions needed for this responsibility |
| Actions | What may be drafted, edited, sent, purchased, or handed back for review |
| Notifications | Recipient, destination, time zone, duration, and changes worth reporting |
| Stop controls | Main task, delegated tasks, and the saved schedule |

Judge the result by the responsibility you assigned. Did it identify the changed deadline, cite the right message, and preserve your sending boundary? Our [coding-agent review guide](/blog/posts/ai-coding-agents) uses the same principle for proposed code: inspect the artifact and evidence. A base model's [benchmark score](/methodology) cannot establish that a Dot will complete this workflow correctly.

The useful next step is to inspect one completed run and its usage before widening the responsibility.

## Questions

### Can I create an OpenAI Dot on my phone?

Create your Dot in the ChatGPT desktop app or a desktop browser first. After setup, the ChatGPT mobile app can reach the same Dot when the supporting update is available. Mobile web is unsupported. Connecting a contact method from the profile is separate from granting access to other apps or your computer.

### Can I use my own API key or choose another model for Dots?

OpenAI documents GPT-6 Astra as the model powering Dots. The launch and setup pages reviewed on September 30, 2026 do not document a Dot model picker or a bring-your-own-key option. Work and Codex have their own model settings. Those settings do not establish a configurable model for the Dot itself.

### Does my computer need to stay on for a Dot to work?

A Dot can keep working in its own cloud computer when your device is off. Work that uses a connected local computer needs that computer online with the ChatGPT app open. Check which environment the task uses before closing the laptop, especially when it depends on local files or installed tools.

### Does disconnecting an app erase what my Dot learned?

Disconnecting an app prevents new access through that connection. Information already added to the Dot's context remains. Deleting the Dot removes its own context, while files, Codex threads, and ChatGPT conversations it created are stored separately. ChatGPT memories also have separate controls, so review each location before assuming the information is gone.