aq.dev / guides / openai-dots-explained

What Is OpenAI Dots? Always-On Agents, Explained

OpenAI Dots is designed to keep each dot's cloud computer and context between tasks, but you should not treat that persistence as your only backup. In November 2026, @PIZURI_DSKB reported that their Dots work data, authentication files, and API-key information disappeared on November 3, and other users reported recurring loss of local files, installed apps, or sign-in state. These are attributed user reports, not evidence of a general outage, a universal cleanup policy, a breach, or exfiltration. OpenAI's current privacy and security FAQ describes context retention and separate file storage, but it does not explain those reports.

OpenAI Dots are always-on personal agents in ChatGPT, announced at DevDay on September 29, 2026. Each dot runs on GPT-6 Astra, OpenAI's flagship model, and is provisioned with its own cloud computer and browser, so it keeps working toward a goal between conversations and after your laptop is closed, then brings results back for your review. You reach the same dot from ChatGPT, Slack, and Microsoft Teams, and it can connect to more than 4,000 apps through OpenAI's plugin ecosystem. For coding, a dot is a coordinator rather than the coder: it creates and steers Codex tasks against your repositories and reports back with the result. This guide explains what OpenAI shipped, the documented persistence and credential boundaries, and the questions a personal agent leaves open for a team.

AQ is the multiplayer coding harness where engineering teams run AI coding agents like Claude Code and Codex together: shared live terminals, a code editor, and app previews, in your own cloud. What is AQ?

What OpenAI announced

OpenAI introduced dots at DevDay 2026 as "remarkably capable, always-on agents built to handle everything." The design is persistent delegation: you give a dot a goal, connect the apps it needs, set what it may do on its own, and it keeps making progress between conversations and brings completed work back for approval, learning your preferences and standards from feedback over time.

Availability, as of November 2026, per OpenAI's announcement and setup guide: dots are rolling out to ChatGPT Pro and Business Premium subscribers, with the first dot included in those plans at no extra cost. The Pro rollout covers users aged 18 and older outside the European Economic Area, Switzerland, and the UK; Business Premium covers supported ChatGPT regions, subject to workspace eligibility. Each member can create one personal dot, while multiple dots per user and teams of dots remain planned. Setup is available on iOS, Android, the desktop app, and desktop web, but not mobile web. Texting is a limited beta for some Pro users in the US, Canada, and Mexico, and is not supported in Business or Enterprise workspaces.

Enterprise access and billing from November 18

OpenAI's Enterprise admin guide says the admin preview for eligible ChatGPT Enterprise, Edu, and Healthcare workspaces runs from November 4 through November 18, 2026. During the preview, an admin must enable dots for a member's role. Starting November 18, members have dots access by default unless the workspace opts out, and each member chooses whether to create one personal dot. The preview is free; billing for enabled dots begins November 18. Eligible Enterprise workspaces can pay a flat rate by work type or based on consumption. OpenAI's setup guide says dots are not available to workspaces with UAE data or inference residency.

These Enterprise dates do not change the personal Pro and Business Premium inclusion above. They also do not change the separate Work and Codex allowances and charges described below.

Admin checks before November 18

Use OpenAI's workspace management guidance and the admin guide to check access, billing, and ongoing work separately:

  1. Review effective access. Check the workspace default and every role assigned directly or through groups. Any one grant can enable dots access, even when another role has it turned off.
  2. Choose the rollout posture before November 18. Decide whether to opt out or run a pilot. For a selected-group pilot, keep the workspace default off and grant access through custom roles.
  3. Read the reports separately. Review the enabled-dot count, which excludes paused dots, alongside active-user counts. Review delegated Work and Codex tasks in those products' own usage and charge reporting.
  4. Test pause, deletion, and offboarding. OpenAI says an admin pause stops applicable billing until the dot returns to active, and deletion pauses billing. Removing or deprovisioning a member revokes access but does not automatically delete their dot, so check the inventory and delete it separately when intended.
  5. Stop independent work explicitly. Pausing a dot does not stop separate Work or Codex tasks or independently scheduled tasks. Review and stop those tasks and schedules separately, and assign an owner for any work that should continue.

What "an agent with its own cloud computer" means

The phrase is literal. Every dot gets a dedicated cloud computer with a browser, separate from your own machine. Three properties follow from that, with important boundaries for persistence and credentials:

  • Persistence is the design, not a backup guarantee. The computer is designed to keep working when yours does not, and OpenAI documents that a dot can retain conversation and plugin context while you keep it. That does not guarantee that every local file, installed app, or browser sign-in will survive between tasks. In November 2026, users reported losing those kinds of local working state, so keep an external copy of important work.
  • Separation by default. Your computer and its contents stay separate unless you explicitly connect them. Access to your local machine is optional and starts turned off, and you can connect only one personal computer at a time.
  • Inspectability, for you. You can open your dot's computer at any time to watch its ongoing work, take over control manually, and hand control back. For supported website sign-ins, OpenAI's protected form sends credentials directly to the browser environment without exposing them to the model. That protection does not cover passwords shared separately in a chat or document, or through a plugin.

What the November reports establish, and what remains unknown

On November 11, 2026, @PIZURI_DSKB wrote that their Dots work data, authentication files, and API-key information disappeared on November 3, and that they had no way to verify what happened or recover it. In a follow-up post, they said another person had reported the same symptoms and argued that the uncertainty was the problem. These posts document one user's account and its stated limits; they do not establish the cause, the number of affected users, malicious deletion, or a breach.

Three additional firsthand accounts describe related symptoms. @Asterisk_Wks reported periodic deletion of the Dots local environment while context carried over, and said they kept a dedicated Google Drive folder. @azurill_yt described suspected roughly daily VM-restart data loss and the resulting backup burden. @natsume_aurlia reported recurring cleanup that removed data and installed apps, Drive backups, repeated reinstalling and re-authentication, and eventually mostly stopping use. These are corroborating user reports, not official confirmation of vendor behavior or measured incidence.

OpenAI's official retention boundaries address different categories of data. A dot can retain its own conversation and plugin context while kept, and deleting the dot starts deletion of that context. Files, Codex threads, and ChatGPT conversations are stored separately and are not deleted with the dot. Files in locations such as ChatGPT Library, your computer, or connected storage follow that location's retention policies. Disconnecting a service stops new access but does not delete information already in the dot's context, and individual dot memories cannot currently be viewed, deleted, or directly modified. These statements distinguish context from separately stored files and conversations; they do not answer every report of a reset in the dot's local environment.

The practical threat model has two separate failure modes: loss of local working state and unintended credential exposure. The reports support concern about the first, but missing API keys or authentication files are not evidence that those credentials were exposed. Assume that a local-environment reset could remove files, installed apps, and browser authentication even when conversation context remains. The cause, scope, and recovery options for the reported resets remain unknown.

Backup and credential checklist

Use checks that cover both failure modes without assuming that a protected sign-in form or a retained conversation protects local files:

  1. Back up important work outside Dots. Export or copy it to storage you control, and restore a sample to confirm that the backup is usable.
  2. Use least-privilege, revocable credentials. Limit each key or token to the services and actions the task needs so you can revoke it without replacing unrelated access.
  3. Rotate keys when their whereabouts are uncertain. Rotation addresses possible exposure; a missing key or authentication file by itself does not show that exposure occurred.
  4. Review connected-plugin permissions. Remove access you no longer need. Remember that disconnecting stops new access but does not erase information already retained in the dot's context.
  5. Verify persistence before relying on it. Create a non-sensitive marker file, install a disposable test app, or sign in a test account, then return in a fresh session after a later task and record what remains. This checks that session only; it does not prove a general retention policy or predict a future reset.

This is the same shift the rest of the agent market made through 2026: the agent's execution environment moves off your laptop and becomes a persistent machine in someone's cloud. For the broader taxonomy, see background agents vs cloud agents; for Cloudflare's version of a persistent agent workspace, see Cloudflare OS agent workspaces, explained.

How a dot does coding work

A dot does not replace Codex; it creates and steers Codex tasks. When you ask your dot for code, it files the work as a Codex task, monitors the result, and follows up: in OpenAI's own example flow, you tell your dot to find the cause of a bug, fix it, add a test that fails without the fix, and open a draft pull request, and the dot comes back with the link.

Mechanically, as of October 2026, that works in two ways. For cloud coding, you first set up a Codex cloud environment naming the GitHub repositories it may use; your dot then creates tasks inside that environment, each running in its own sandbox preloaded with the repository. For local coding, a dot can create or continue Codex tasks on a computer you have connected. Repository access goes through the GitHub connector, which lets the agent investigate issues and prepare pull requests in the repositories you granted.

Three accounting details matter for anyone budgeting agent usage. Talking to your dot does not count toward your ChatGPT usage limits, but the tasks it starts in Codex or ChatGPT Work count toward those products' limits, so a busy dot spends your Codex allowance. And OpenAI describes "an allowance for deeper work, with extended limits for the first month after launch," without published post-launch numbers. For Enterprise workspaces, OpenAI says tasks delegated to Work or Codex incur separate standard charges and appear in those products' reporting, so review them alongside the enabled-dot count rather than as part of it. For what Codex tasks look like once created, see Codex background agents and the agents dashboard.

Permissions: what a dot may do on its own

OpenAI's approval rules distinguish actions a dot may take on its own, actions that need approval, and actions that require you to take over. The most sensitive actions, like changing a password or transferring money, require you to complete them personally. Others, like permanently deleting data or installing software, may require approval each time, and some recurring actions can be approved in advance. Custom Rules let you set boundaries for supported actions, but they cannot switch off core safety requirements, Auto-review, or the restrictions on proactive research. Auto-review checks certain planned actions against your instructions, Custom Rules, and safety requirements before they run.

On data: a dot can receive memories and recent conversation context from ChatGPT, and conversations with it can contribute back to ChatGPT memory, but OpenAI states the dot's context does not retain credentials, images, or screenshots. That is a statement about retained context, not blanket coverage for credentials supplied outside supported sign-in flows or a guarantee about local authentication files. The same FAQ says prompt-injection protections reduce risk without eliminating it, dots can still make mistakes, consequential work needs review, and whether a completed action can be reversed depends on the action and app involved.

Who else can see or drive the session?

Here is the open question for engineering teams. A dot is a personal agent: one person's goals, app connections, approval queue, and cloud computer. ChatGPT Space, announced at the same DevDay, gives teams and agents a shared place for files and project context, and business workspaces get admin controls for managing dots. But the working session itself belongs to its owner: a teammate cannot open your dot's computer, watch the diff take shape, or take the keyboard when the agent heads down the wrong path.

That matters because solo-by-default is already the pattern the data shows. A July 2026 LeadDev analysis of 25,264 agent-generated pull requests across 2,361 popular GitHub repositories found that in 79 percent of agentic PRs the same developer both reviewed and modified the agent's contribution, and only about one in eight workflows involved multiple humans. Personal always-on agents extend what one person can delegate; they do not, by themselves, make agent work visible to the team that has to review, debug, and live with it.

TopicA dot's cloud computerA team coding workspace
Belongs toOne person, inside their ChatGPT planThe team, on infrastructure it chooses
The coding agentCodex tasks the dot creates and steersWhichever CLI the engineer runs: Claude Code, Codex, Cursor Agent, others
Who can watch liveThe owner, by opening the dot's computerTeammates in the same workspace, in the browser
Who can take the keyboardThe owner (take over, return control)The session owner, plus anyone they approve
Where code livesCodex sandboxes per task, repo via the GitHub connectorA git worktree per workspace on a persistent machine

Neither column is wrong; they answer different questions. A dot asks how much of one person's work an agent can carry. A team workspace asks how a group ships software with agents it can watch, steer, and review together.

Where AQ fits

AQ is the multiplayer coding harness where engineering teams run AI coding agents like Claude Code and Codex together: shared live terminals, a code editor, and app previews, in your own cloud. In this guide's terms, AQ is what the always-on cloud computer looks like when it is built for a team instead of a person. Agents run as the real CLIs (Claude Code, Codex, Cursor Agent, Kimi, Grok, or plain shells) in persistent tmux sessions on your team's VM, streamed live to the browser: sessions survive a closed laptop, resume from any device, and teammates open the same workspace and watch the same live session, with typing into someone else's terminal delegated by its owner approving a control request in one click. Each workspace gets its own isolated git worktree with automatic dependency install, agents commit, push, and open PRs with per-user GitHub auth, and every workspace serves a live dev-server preview with shareable links. Visibility is owner-managed: a workspace is team-visible, or private and shared with specific people. Engineers sign into the CLIs with their own Claude and OpenAI accounts (AQ never marks up usage on your own subscriptions), and execution is either VMs you connect from your own cloud or a dedicated always-on AQ-managed VM in its own isolated network, with no shared multi-tenant execution tier. The Free plan runs agents on your own Mac with the AQ Mac app, with no time limit; the AQ Pro plan is $50 per user per month in early access (standard $200, billed monthly), rate locked for your first 12 months.

Plainly: a dot is a personal chief of staff that happens to file Codex tasks; a coding harness is where the engineering team does the coding work in the open. The optional Dots importer brings bot instructions, routines, and memories you choose into AQ, with routines paused until you turn them on and passwords, keys, and tokens excluded. It does not import Dots files, local application state, or credentials, and it cannot recover lost Dots data. If your team is already running Codex in the cloud, running Codex on a cloud VM and keeping Codex running after closing the laptop cover the workspace side of the same shift.

Frequently asked questions

What is OpenAI Dots?

Dots are always-on personal agents in ChatGPT, announced at OpenAI's DevDay on September 29, 2026. Each dot runs on GPT-6 Astra, has its own cloud computer and browser, keeps working toward your goals between conversations, connects to more than 4,000 apps, and brings completed work back for your review. You can message the same dot from ChatGPT, Slack, and Microsoft Teams.

Do Dots files, installed apps, and browser sessions always persist?

Persistence is the documented design, not a backup guarantee. OpenAI says a dot can retain conversation and plugin context while kept, and that files, Codex threads, and ChatGPT conversations are stored separately. In November 2026, four users reported losing local files, installed apps, or authentication state. Those reports do not establish the cause or incidence, but they are a reason to keep important work in an external backup and verify a fresh session before relying on persistence.

Do Dots protected sign-in forms keep every credential away from the model?

No. OpenAI's protected form applies to supported sign-in flows and sends those credentials directly to the browser environment without exposing them to the model. It does not cover passwords shared separately in chat or documents, or through a plugin. OpenAI also says dot context does not retain credentials, but that does not make every credential-handling path a protected sign-in. Missing API keys or authentication files are not evidence of exposure; use revocable least-privilege credentials and rotate keys when their whereabouts are uncertain.

Which ChatGPT plans include a dot?

As of November 2026, OpenAI says the first dot is included at no extra cost with ChatGPT Pro and Business Premium, rolling out gradually. Pro covers users aged 18 and older outside the European Economic Area, Switzerland, and the UK; Business Premium covers supported ChatGPT regions, subject to workspace eligibility. Eligible Enterprise, Edu, and Healthcare workspaces have separate access and billing rules from November 18. OpenAI says support for multiple dots per user is planned.

When does Enterprise Dots billing start?

OpenAI's admin guide says the admin preview is free through November 18, 2026, when billing for enabled dots begins. Members get dots access by default from that date unless the workspace opts out. Eligible Enterprise workspaces can pay a flat rate by work type or based on consumption. Delegated Work and Codex tasks incur separate standard charges and appear in those products' reporting.

Can an OpenAI dot write code?

Indirectly. A dot coordinates coding work rather than doing it: it creates and steers Codex tasks, either in a Codex cloud environment you set up with named GitHub repositories, or on a computer you have connected. Each cloud task runs in its own sandbox preloaded with the repository, and the dot can bring back a draft pull request. Tasks the dot starts in Codex count toward Codex usage limits, even though talking to the dot does not count toward ChatGPT limits.

Can my teammates see or drive my dot's session?

No. A dot is personal: its goals, app connections, approvals, and cloud computer belong to one ChatGPT account, and only the owner can open its computer or take over control. ChatGPT Space gives teams a shared place for files and project context, and business workspace admins can manage dots, but the live working session is not something a teammate can watch or steer.

Is a dot the same thing as Codex?

No. Codex is OpenAI's software engineering agent: it writes code, runs tests, and proposes pull requests, each task in its own sandbox. A dot is a general always-on agent that delegates to Codex for coding work, monitors the result, and reports back. OpenAI positions the dot as the coordinator and Codex as the worker; using one does not replace the other.