OpenAI Dots Autonomous Agent: Setup Guide & Key Features - Features

OpenAI Dots Autonomous Agent: Setup Guide & Key Features

Learn how the OpenAI Dots autonomous agent works: GPT-6 Astra, dedicated cloud computers, connected apps, scheduled tasks, and permissions.

2026-09-30
OpenAI Dots Wiki Team
Quick Guide
  • OpenAI Dots autonomous agent launched at DevDay on September 29, 2026, powered by GPT-6 Astra.
  • Each Dot runs on a dedicated cloud computer with its own browser for long-running tasks.
  • Dots connect to 4,000+ apps and follow up via ChatGPT, email, and Slack.
  • Permissions, autonomy settings, and workspace controls define what a Dot can access.
  • Best use cases: ongoing research, scheduled monitoring, reminders, and multi-step workflows.

What Is the OpenAI Dots Autonomous Agent?

The OpenAI Dots autonomous agent is an always-on AI assistant introduced at OpenAI's DevDay on September 29, 2026. Unlike a standard ChatGPT conversation that ends when you close the tab, a Dot is a persistent agent that keeps working toward an assigned goal across multiple sessions. It is powered by GPT-6 Astra, OpenAI's flagship model designed for complex reasoning, coding, computer use, and research tasks.

What separates Dots from ordinary chat sessions comes down to three pillars:

Always-On Design

  • Persistent goals instead of one-off prompts
  • Continues working between conversations
  • Surfaces results proactively via ChatGPT, email, or Slack

Dedicated Cloud Computer

  • Each Dot gets its own cloud PC and browser
  • Executes multi-step web tasks independently
  • Keeps running when you close your local browser

Connected Apps

  • Access to 4,000+ app connections
  • Works across permitted services and channels
  • Every connection follows user or workspace permissions
Editor's Insight

Think of a Dot as a delegation target, not a chat partner. The best results come from giving it a clear, ongoing outcome — "monitor this market weekly" — rather than a single isolated request.

The table below summarizes the core facts confirmed in OpenAI's launch materials:

AttributeDetail
Release dateSeptember 29, 2026 (DevDay)
Underlying modelGPT-6 Astra
Execution environmentDedicated cloud computer with browser
App connections4,000+ supported tools
Follow-up channelsChatGPT, SMS, email, Slack
Safety documentationDedicated Dots section in official OpenAI safety docs

How to Set Up Your First Dot

Creating a Dot takes just a few minutes, but the configuration choices you make determine how well it performs over the following weeks. Follow this workflow to build a Dot that stays useful:

1

Start a New Dot

Open the Dots experience in ChatGPT on desktop web or the supported desktop app and choose to create a new dot.

2

Choose a Descriptive Name

Name the Dot after its responsibility — a project, research topic, or recurring operational task — so it is easy to manage alongside other dots.

3

Define the Goal

Write a clear, ongoing outcome the Dot should continually work toward. A focused goal gives the agent stable direction across work sessions.

4

Configure Autonomy and Boundaries

Set instructions for what the Dot should prioritize, when it may act independently, and when it should ask for approval before sensitive actions.

5

Attach Context and Files

Add relevant documents and reference material so the Dot has the project context it needs from day one.

6

Review the Activity Dashboard

Open the Dot's activity view to inspect completed actions, follow current tasks, and adjust its configuration as work progresses.

Avoid Vague Goals

A goal like "help me with email" produces scattered behavior. Specify scope, cadence, and boundaries — for example, "draft replies for inbound customer email each morning, flag anything about refunds for my review."

The Dedicated Cloud Computer

The cloud computer is the feature that most clearly distinguishes the OpenAI Dots autonomous agent from a normal ChatGPT conversation. Every Dot operates inside its own remote computer environment, complete with a browser, allowing it to perform multi-step web work that continues independently of your local machine.

CapabilityWhat It Means for You
Dedicated cloud environmentThe Dot works in its own cloud PC, not your browser tab
Independent executionLong tasks keep running while you do other things
Computer viewInspect the current state of the Dot's work at any time
Manual interventionStep into the session when review or input is needed
Cloud vs. local accessOptional local-computer access is separate from cloud execution
Cloud vs. Local Access

The dedicated cloud computer is the default working environment. Access to your own local computer is an optional, separately controlled capability — keeping remote agent execution distinct from actions on your personal device.

Apps, Scheduled Tasks, and Use Cases

Connected apps turn a Dot from a clever chatbot into a working assistant. Connections follow the permissions granted by the user or workspace, and the Dot can only act within those boundaries.

Ongoing Research

  • Assign an evolving topic
  • Dot collects and organizes findings over time
  • Updates across multiple work sessions

Recurring Monitoring

  • Repeatedly checks changing information
  • Surfaces meaningful updates proactively
  • Ideal for market, news, or status tracking

Cross-App Workflows

  • Combines actions across permitted apps
  • Handles multi-step agentic work end to end
  • Follows up through your chosen channels

Scheduled tasks are the backbone of always-on value. A Dot can run work at a specific future time or on a recurring schedule, then deliver proactive updates without a new prompt:

Task TypeExampleCadence
ReminderSurface a deadline follow-upOne-time
Recurring checkReview connected data for changesDaily / weekly
Recurring summaryPrepare a digest of the week's activityWeekly
Event-triggered follow-upAct when new messages arriveContinuous
Best Practice

Start with one scheduled task per Dot. Reviewing active, scheduled, and completed work from the task management view each week keeps autonomy predictable and easy to audit.

Permissions, Availability, and Controls

Permissions define the boundaries within which any OpenAI Dots autonomous agent operates. Both individual users and workspace administrators have meaningful control.

ControlScopeDescription
Connected app permissionsUser and workspaceDetermines which services a Dot may access
Cloud computer accessDot capabilityEnables browser-based multi-step tasks
Local computer accessOptional user capabilitySeparately granted access to your own device
Autonomy settingsUser configurationHow independently the Dot acts between interactions
Workspace availabilityAdministratorAdmins decide whether members can use Dots
Enterprise access controlsEnterprise adminOrganization-level restrictions and governance

Availability at a glance:

Platform / AccessStatus
Desktop webSupported for creation and management
Desktop appSupported for users in the rollout
MobileRollout-dependent; advanced config centered on desktop
Plan accessTied to supported ChatGPT plans and current rollout
EnterpriseControlled rollout with administrator governance
Check Your Workspace

In Business and Enterprise workspaces, administrator settings determine whether Dots are available at all and which connected tools members may use. Confirm availability before planning workflows around a Dot.

Before You Delegate Work to a Dot:

  • Confirm your plan and workspace allow Dots access
  • Write a focused, ongoing goal — not a one-off prompt
  • Set autonomy boundaries and approval rules
  • Connect only the apps the workflow requires
  • Review the activity dashboard on a regular schedule

FAQ

Q: What is the OpenAI Dots autonomous agent?

Dots are persistent, always-on AI assistants announced at DevDay on September 29, 2026. Powered by GPT-6 Astra, each Dot has its own cloud computer and browser, connects to apps, and continues working on long-running tasks between conversations.

Q: How is a Dot different from a normal ChatGPT conversation?

A conversation ends when you stop typing. A Dot is organized around persistent goals, can run scheduled and multi-step work in its dedicated cloud computer, and proactively follows up through channels like ChatGPT, email, and Slack.

Q: Can I control what my Dot is allowed to access?

Yes. Connected-app permissions, computer capabilities, autonomy settings, and workspace-level administrator controls all define what a Dot can do. Optional local-computer access is granted separately from cloud execution.

Q: Where can I create and manage Dots?

Users with access can create and configure Dots through ChatGPT on desktop web or supported desktop experiences. Mobile access varies by account and rollout stage, with advanced configuration centered on desktop.