OpenAI dots
OpenAI's personal assistant that uses plugins and shared ChatGPT memory.
Group: Personal agents you message
OpenAI dots uses shared ChatGPT/plugin permissions and memory for connected work, proactive private research, and permitted local or cloud computer tasks. Its FAQ distinguishes read-only proactive research, action approvals, Custom Rules and takeover, while Enterprise workspace controls separately govern access and messaging.
The decision in plain English
Check plan, region and the 18 November transition before choosing dots. Current OpenAI docs list Pro rollout outside the EEA, Switzerland and UK, Business Premium in supported regions subject to workspace eligibility, and eligible Enterprise/Edu/Healthcare access during a 4-18 November admin preview. Enterprise preview is free through 18 November; billing starts then for enabled dots, with delegated Work/Codex billed separately. We have not verified exact dollar rates or account signup eligibility. Access, action permissions, retained data and independent tasks are separate controls.
Best fit
Personal connected-app work with shared ChatGPT context.
Poor fit
A user needing per-memory inspection and deletion.
Documented capabilities
Vendor statements, not verified task outcomes.
- Shared connections: Plugin permissions apply across dots, ChatGPT, Work and Codex; enabling dots does not grant all apps. Manage actual plugin scopes separately. Source · checked 2026-10-09
- Proactive research: Permitted sources can feed private notes, but research tools cannot directly send messages, change plugin content or control browser/computer. Follow-up actions use normal authorization. Source · checked 2026-10-09
- Local device access: Connect the device and grant camera/mic/screen access in device settings. Enterprise local command/file access additionally needs admin enablement; app must be open and device online. Source · checked 2026-10-09
- Cloud computer: Browser, network, desktop and password-manager capabilities are separately managed. Cloud does not inherit local VPN, sign-ins or device policies. Source · checked 2026-10-09
- Rules and approvals: Custom Rules support boundaries but cannot remove core safeguards/Autoreview. Sensitive transfers/password changes require takeover; some messages and purchases accept specifically scoped advance approval. Source · checked 2026-10-09
- Messaging channels: the desktop app connects messaging channels such as Slack or texting; dots are not supported on mobile web. Source · checked 2026-10-29
Permissions and human control
- Plugin permissions are shared across dots, ChatGPT, ChatGPT Work and Codex.
- Custom Rules add boundaries but cannot override core safeguards. Sensitive actions can require takeover or approval; scoped advance approval is supported for some actions.
- Enterprise access is off by default during the admin preview through 18 November, then available by default unless admin opts out; app authorization and local/cloud permissions are separate.
- Only owner instructions direct an Enterprise dot; other people's Slack DMs/mentions do not start work. Posts in channels are visible to members.
- Proactive research tools cannot directly message others, write through plugins or control browser/computer. A follow-up action still needs normal authorization; Custom Rules cannot remove this restriction.
- Password changes and money transfers require takeover. Saved-card purchases and recurring messages may use narrowly scoped advance approval, not open-ended ongoing authority.
Documented controls do not prove that an agent always follows them.
Limits and open questions
- Individual dot memories cannot currently be viewed or edited
- Disconnecting services does not delete retained context
- Phone messaging is unavailable for Enterprise at launch, according to workspace docs.
Privacy and security
OpenAI says content is encrypted in transit and at rest. Disconnecting stops new plugin access, not deletion of retained context. Delete the dot to delete its context; separately stored files, conversations and ChatGPT Memory remain separate. Individual dot memories cannot currently be viewed or edited. Business, Enterprise and Edu data is not used for training by default; personal plans use the Improve the model for everyone setting. Human safety review can still occur. Vendor statements, not an audit. Proactive notes are not directly used for training, but information brought into eligible conversations/tasks may be, depending on settings. Turning off shared Memory does not erase received context. Supported secure-login forms differ from passwords sent in chat/plugin content.
Feature-by-feature: what is documented and what is not
Official sources checked 23 October 2026. A documented feature is a vendor statement, not proof of success.
| Feature | Evidence status | Detail and limit |
|---|---|---|
| Shared connections | Documented | Plugin permissions apply across dots, ChatGPT, Work and Codex; enabling dots does not grant all apps. Manage actual plugin scopes separately. Source |
| Proactive research | Read/notes only | Permitted sources can feed private notes, but research tools cannot directly send messages, change plugin content or control browser/computer. Follow-up actions use normal authorization. Source |
| Local device access | Separate permission | Connect the device and grant camera/mic/screen access in device settings. Enterprise local command/file access additionally needs admin enablement; app must be open and device online. Source |
| Cloud computer | Distinct controls | Browser, network, desktop and password-manager capabilities are separately managed. Cloud does not inherit local VPN, sign-ins or device policies. Source |
| Rules and approvals | Not universal ask-first | Custom Rules support boundaries but cannot remove core safeguards/Autoreview. Sensitive transfers/password changes require takeover; some messages and purchases accept specifically scoped advance approval. Source |
| Memory | Deletion distinction | Shares ChatGPT memories/context; turning Memory off stops sharing, not existing context. Individual dot memories cannot be inspected/edited/deleted individually. Source |
| Delete/disconnect | Different effects | Disconnect stops new access. Delete dot removes its own context; files, Codex threads, ChatGPT conversations and Memory remain separate. Source |
| Training | Plan/setting-dependent | Personal plans use Improve the model for everyone; Business/Enterprise/Edu not trained by default. Proactive notes are not directly trained on, but their content in eligible work may be. Source |
| Work messaging | Enterprise caveat | Supported Slack/Teams setup needs admin grants and member connection; Slack may require app approval. Only owner directs dot; others' mentions/DMs do not start work. Phone iMessage/RCS/WhatsApp unavailable at Enterprise launch. Source |
| Credentials and purchases | Supported-flow limits | Secure login form keeps supported credentials from the model; chat/plugin passwords are not covered. Auth checks may need takeover; saved-merchant-card purchases require scoped approval. Source |
| Price and eligibility | Plan/region dependent | Pro and Business Premium access is plan/region dependent; eligible Enterprise free preview through 18 November, then billing for enabled dots. Delegated Work/Codex separate. Exact dollar rates unverified. Source |
How to start with OpenAI dots narrowly
- Confirm your actual account has dots access; published Enterprise guidance is not proof of personal-plan eligibility or a free standalone service.
- Workspace admins: review effective roles and Use dots access before the 18 November default-enable and Enterprise billing transition, plus local/cloud permissions and app authorizations. Do not assume enabled access creates a dot or disabling access stops every independent task.
- Review shared plugin access in ChatGPT and set narrow Custom Rules before delegating a task. Select a draft/read-only task first.
- If using a computer, connect it and grant only necessary device permissions; distinguish local access from cloud sessions and their separate sign-ins.
- Review Activity View and stop/change instructions as needed. Before removing access, separately consider dot context, stored files/conversations and ChatGPT Memory; disconnecting alone is not deletion.
Official product/setup source. These are documentation-based steps, not a hands-on setup test.
Common questions about OpenAI dots
What is OpenAI dots for?
Personal connected-app work with shared ChatGPT context. The vendor describes it as: OpenAI dots uses shared ChatGPT/plugin permissions and memory for connected work, proactive private research, and permitted local or cloud computer tasks. Its FAQ distinguishes read-only proactive research, action approvals, Custom Rules and takeover, while Enterprise workspace controls separately govern access and messaging.
Who should look elsewhere?
A user needing per-memory inspection and deletion.
How much does OpenAI dots cost?
Current getting-started docs describe eligible Enterprise flat-rate or scaled-consumption billing. Free preview 4-18 November 2026; billing starts 18 November for enabled dots. Work/Codex delegated threads billed separately at standard rates. No exact dollar rates established here. Pro and Business Premium access is listed separately; no universal free standalone service claimed.
What can it access and what needs my approval?
Plugin permissions are shared across dots, ChatGPT, ChatGPT Work and Codex. Custom Rules add boundaries but cannot override core safeguards. Sensitive actions can require takeover or approval; scoped advance approval is supported for some actions. Enterprise access is off by default during the admin preview through 18 November, then available by default unless admin opts out; app authorization and local/cloud permissions are separate. Only owner instructions direct an Enterprise dot; other people's Slack DMs/mentions do not start work. Posts in channels are visible to members. Proactive research tools cannot directly message others, write through plugins or control browser/computer. A follow-up action still needs normal authorization; Custom Rules cannot remove this restriction. Password changes and money transfers require takeover. Saved-card purchases and recurring messages may use narrowly scoped advance approval, not open-ended ongoing authority.
What happens to my data?
OpenAI says content is encrypted in transit and at rest. Disconnecting stops new plugin access, not deletion of retained context. Delete the dot to delete its context; separately stored files, conversations and ChatGPT Memory remain separate. Individual dot memories cannot currently be viewed or edited. Business, Enterprise and Edu data is not used for training by default; personal plans use the Improve the model for everyone setting. Human safety review can still occur. Vendor statements, not an audit. Proactive notes are not directly used for training, but information brought into eligible conversations/tasks may be, depending on settings. Turning off shared Memory does not erase received context. Supported secure-login forms differ from passwords sent in chat/plugin content.
What changes for workplace dots on 18 November 2026?
OpenAI documents eligible Enterprise/Edu/Healthcare admin preview until 18 November, then default availability unless admins opt out. Its getting-started page says Enterprise billing begins on that date for enabled dots; delegated Work/Codex usage is separate. Admin pause or access removal does not delete data or prove all independent work stopped. Review effective roles, cost tiers, separate tasks and residency eligibility. Vendor documentation, not our test.
Has AgentComparison tested it?
No. This record is built from the vendor's own documentation, checked 2026-10-23. Vendor claims are attributed, not confirmed.
How to evaluate OpenAI dots yourself
- Pick one low-risk task that matches its documented capabilities: shared connections: plugin permissions apply across dots, chatgpt, work and codex; enabling dots does not grant all apps. manage actual plugin scopes separately..
- Write the expected output and what counts as failure before you start.
- Connect the minimum access needed, and check approval settings first.
- Run it twice and compare, then review every action it took.
- Check cost after any free allowance. Unknown price is not free.
Related
Documented use cases: Personal email, calendar and reminders.
- Instinct (product): A messaging-first assistant connected to personal apps and devices.
- Poke (product): Messaging-based email, calendar and reminder work.
- Lucas (product): Personal administration through iMessage.
- Muse (product): Personal reminders, ongoing tasks and artifact work.
- Cue by Manus (product): People who want several agents with separate identities that hand work to each other, and who will review the final result.
- Instinct vs Muse vs OpenAI dots
- AI shopping agents compared for retailers
- AI agent use cases by job
- Privacy and approval checklist
- Compare two agents side by side
- Latest agent updates
Sources and history
- https://help.openai.com/en/articles/20001529-dots-privacy-security-and-safety-faqs · official documentation or policy, read 9 October 2026 · read 23 October 2026
- https://help.openai.com/en/articles/20001554-manage-dots-in-chatgpt-workspaces · official documentation or policy, read 9 October 2026 · read 23 October 2026
- https://help.openai.com/en/articles/20001530-getting-started-with-your-dot · official · read 23 October 2026
- https://learn.chatgpt.com/docs/whats-new/dots-october-28-2026 · official · read 23 October 2026
- https://help.openai.com/en/articles/20001554-managing-dots-in-chatgpt-workspaces · official · read 23 October 2026
- https://help.openai.com/en/articles/10128477-chatgpt-enterprise-and-edu-release-notes · official · read 23 October 2026
Change history
- 2026-10-09: Re-read official features, permissions, pricing and privacy sources; added product-specific evidence table and narrow-start steps. Vendor documentation, no hands-on tests.
- 2026-10-07: Rechecked public FAQ and workspace guidance; clarified training settings, deletion and approval scope.
- 2026-10-01: Initial documentation research. No hands-on testing.
- 2026-10-23: undefined
- 2026-10-29: Field-level audit: OpenAI getting-started help article (vendor-updated 28 Oct) read - rollout documented for Pro 18+ (excl. EEA/CH/UK) and Business Premium (supported regions), Enterprise beta admin-enabled; first dot included in plan at no extra cost with deeper-work allowance; messaging channels (Slack, texting) via desktop; availability/pricing/capabilities/meta updated. checked_label unchanged (not a full-record check).
- 2026-11-01: Texting channel added: OpenAI's 'Text your dot' doc (week of 26-30 October) announces iMessage/RCS texting in beta for ChatGPT Pro users in the US, with text, photos and voice messages; channel chosen by context or on request; approvals via link into ChatGPT. checked_label stays 2026-10-23 (single-feature re-check only).
- 2026-11-01: Decision note rewritten as a buying decision (plan and region framing) from the vendor getting-started article and the texting beta doc; disconnection caveat retained. Documentation research only.
- 2026-11-05: Reconciled current and legacy workspace docs against 4 November release notes: admin preview/default access transition, Enterprise billing boundary, independent task and data controls. Documentation only; no signup or paid test.
Documentation overview expanded with current pricing/access sources. No hands-on results claimed.
Report a correction