OpenAI Dots

OpenAI Dots are always-on personal AI agents powered by GPT-6 Astra that use cloud computers, connected apps, memory, proactive research, and scheduled tasks to complete ongoing work.

About OpenAI Dots

OpenAI Dots are always-on AI agents developed by OpenAI and introduced on September 29, 2026. Powered by GPT-6 Astra, each dot has its own cloud computer and browser, allowing it to research information, create deliverables, and work toward assigned goals. Dots are designed to take ongoing responsibility for projects rather than operate only through individual conversations.

Users can assign goals, provide instructions, connect applications, and define how independently their dot can act. A dot can continue progressing between conversations, handle multiple projects, and review connected information proactively. It can also perform scheduled tasks, identify relevant updates, and use information from ChatGPT memory and connected applications to maintain context.

Dots can interact with applications through OpenAI's plugin ecosystem, which includes more than 4,000 supported applications. They can also delegate work to ChatGPT Work and Codex, interact with cloud development environments, and use a connected local computer when explicitly authorized. Users can communicate with their dot through ChatGPT, supported messaging channels, and voice conversations.

OpenAI Dots include configurable action permissions, human approval requirements, activity monitoring, and security safeguards. Users can inspect the agent's computer, review task progress, provide corrections, and pause execution. The product is intended for professionals, developers, researchers, content creators, and organizations managing projects that require continuing attention and coordination across applications.

Key Features

- Autonomous Cloud Computer: Each dot operates through its own cloud computer and browser, enabling independent research, application interaction, document creation, and task execution. Users can inspect the computer and take control when necessary.

- Proactive Research and Scheduled Tasks: Dots can review permitted connected information in the background, identify relevant updates, and create private research notes. Users can also schedule reminders, recurring checks, and other tasks that run without a new conversation. Proactive research tools have read-only restrictions.

- Persistent Memory and Context: Dots receive relevant ChatGPT memories and conversation context, create memories from ongoing interactions and connected applications, and use this information to understand goals, preferences, and previous work.

- Application Integrations and Agent Delegation: Dots can access more than 4,000 applications through OpenAI's plugin ecosystem. They can use supported connected applications and delegate tasks to ChatGPT Work, Codex, and configured Codex cloud environments.

- Configurable Actions and Human Approval: Custom Rules let users specify whether supported actions can proceed independently, require prior authorization, need confirmation, or must be handed back to the user. Built-in safeguards apply regardless of custom permissions.

- Activity Monitoring and Security Controls: Activity View displays ongoing and delegated tasks, execution status, and completed work. Auto-review evaluates certain planned actions against instructions and safety requirements, while additional safeguards address prompt injection and potentially harmful activity.

Pricing

ChatGPT Pro: Included with an eligible ChatGPT Pro subscription at no additional charge for the first dot. Includes an allowance for deeper agent work and extended limits during the initial launch period.
ChatGPT Business Premium: Included with an eligible Business Premium subscription at no additional charge for the first dot. Availability depends on supported regions and workspace eligibility.
ChatGPT Enterprise, Edu and Healthcare: Beta access is available when enabled by an eligible workspace administrator. Access is initially disabled by default.

Additional Usage Information:

Conversations with a dot do not count toward normal ChatGPT usage limits.
Tasks delegated to ChatGPT Work or Codex consume their respective usage allowances.
The included agent work allowance has extended limits during the first month after launch.
Additional dots and increased execution capacity are planned for future availability.
Separate public charges for additional dots are not specified.

Use Cases

Personal Productivity and Scheduling: Review connected calendars, monitor upcoming commitments, schedule recurring checks, prepare reminders, and identify tasks requiring attention.
Software Development and Testing: Monitor customer feedback, investigate bugs, implement smaller improvements, test code changes, and prepare pull requests for developer review.
Content Creation and Marketing: Process interview transcripts, prepare show notes, draft social media content, update campaign materials, and revise documents when project requirements change.
Business Research and Project Management: Analyze new information, update research findings, prepare presentations, review customer requirements, develop proofs of concept, and maintain progress on ongoing projects.

Pros & Cons

Pros:

Provides an independent cloud computer and browser, supports multiple ongoing projects, and continues working between conversations.
Combines persistent personal context, proactive research, recurring task execution, and access to more than 4,000 applications through plugins.
Supports task delegation to ChatGPT Work and Codex, configurable action approvals, activity inspection, and automated security reviews.

Cons:

Availability is restricted by subscription, region, age, and workspace eligibility. Mobile web is not supported, and Enterprise access requires administrator enablement.
Local computer execution requires explicit permission, a connected computer, and the ChatGPT desktop application running while the computer is online. Some sensitive actions require user takeover.
Individual dot memories cannot currently be viewed, modified, or deleted separately. Resetting the dot is required to delete its own retained context.

Integrations

ChatGPT, ChatGPT Work, OpenAI Codex, Codex Cloud Environments, ChatGPT Plugins, Slack, Microsoft Teams, Google Drive, Gmail, GitHub, ChatGPT Memory, ChatGPT Library, ChatGPT Desktop App, ChatGPT iOS App, ChatGPT Android App, Local Computer, Cloud Browser, Microsoft Agent 365 (specialist dots integration in development)

FAQ

What are OpenAI Dots and how do they work?

OpenAI Dots are always-on personal AI agents powered by GPT-6 Astra. Each dot has its own cloud computer and browser and can use connected applications to research, create documents, develop software, and complete assigned work. Users provide goals and instructions, while the dot determines the steps needed and continues making progress between conversations.

Can OpenAI Dots work automatically without receiving a new message?

Yes. Dots can conduct proactive background research using permitted connected information, identify useful updates, and run scheduled reminders or recurring tasks. Users can also assign longer projects that continue between conversations. Proactive research tools are restricted to read-only operations and cannot directly send messages, modify connected content, or control computers.

Which applications and development tools can OpenAI Dots access?

Dots can use OpenAI's plugin ecosystem, which supports more than 4,000 applications. They share eligible plugin connections with ChatGPT, ChatGPT Work, and Codex. Officially documented systems include Slack, Microsoft Teams, Google Drive, Gmail, GitHub, Codex cloud environments, and optionally connected local computers. Actual access depends on permissions, account connections, and workspace settings.

Do OpenAI Dots remember previous conversations and user preferences?

Yes. Dots receive relevant memories and conversation context from ChatGPT and can create their own memories from interactions and connected applications. They use this context to support ongoing projects and adapt to user preferences. Individual memories stored in a dot's own context cannot currently be directly edited or deleted separately.

Can OpenAI Dots execute actions without human approval?

Dots can perform permitted actions independently when their instructions and approval settings allow it. Custom Rules control supported actions that can proceed automatically, require preapproval, need confirmation, or must be handed back to the user. Sensitive activities such as password changes and transferring money require user takeover. Auto-review and built-in safety restrictions cannot be disabled through Custom Rules.

How can users monitor and control an OpenAI Dot?

Users can review ongoing, scheduled, and completed tasks through the dot's profile and Activity View. They can inspect its cloud computer, review execution steps, provide corrections, modify instructions, change task schedules, or pause the dot. Local computer access can also be revoked. Dots use automated action reviews and security monitoring, although consequential results should still be checked by the user.

Categories:

AI Agents Personal AI Agents