How to Use ChatGPT Dots

Learning how to use ChatGPT dots starts by creating your agent in the ChatGPT desktop application or on desktop web, configuring permissions, and assigning work across your accounts. OpenAI introduced dots on September 29, 2026, as always-on agents powered by GPT-6 Astra that execute tasks independently in the background.

Each dot runs on a dedicated cloud computer with its own browser, allowing it to complete assignments without requiring an open chat window. Because OpenAI launched dots today, broad performance benchmarks are not yet available, and account access continues to roll out across supported paid tiers.

OpenAI says dots can still make mistakes, so review consequential work before you act on it.

Five Steps to Set Up and Configure Your Dot

Setting up a dot requires creating the agent on a desktop browser or desktop application, choosing an identity, granting tool access, and defining approval thresholds.

OpenAI limits initial setup to desktop environments. You cannot create a dot on mobile devices, and mobile web browsers do not support the setup process. Once configured on desktop, your dot syncs across channels, allowing you to monitor tasks or send instructions from mobile applications and connected workplace chat platforms.

• Verify subscription tier and regional access: Ensure you have an active ChatGPT Pro account outside the European Economic Area, Switzerland, and the UK, or a ChatGPT Business workspace with a Premium seat.
• Launch setup on desktop: Open the ChatGPT desktop application on macOS or Windows, or visit ChatGPT on desktop web, and follow the onboarding prompt to create your primary dot.
• Select an identity and handle: Assign a display name, choose an avatar or allow the dot to generate a pet illustration, and confirm your handle (@yourname-agentname or the default @yourname-dot).
• Connect external software: Connect apps from the over 4,000 that OpenAI's plugin ecosystem supports, reviewing access permissions before approving each connection.
• Configure Custom Rules and assign a first task: Specify which actions require explicit human approval, then submit an initial research or tracking assignment.

How to Use ChatGPT Dots for Daily Tasks and Delegated Work

Delegating work to a dot involves describing your goal, attaching reference files with the plus icon, and stating whether the task runs once or recurs on a schedule.

Your dot manages projects independently and can handle multiple initiatives at the same time without requiring separate conversation threads. You can monitor active operations by opening the dot profile, where work is organized under In progress, Scheduled, and Completed tabs. Scheduled tasks let you establish ongoing checks, such as scanning calendar events each morning or tracking specific project updates, with controls to edit intervals or notification settings.

You can open your dot's cloud computer at any time from its desktop profile to inspect its work. On mobile devices, the virtual computer view opens under your direct control if manual intervention is required. When a project requires deep coding or long-form documentation, your dot can initiate tasks in ChatGPT Work and Codex, which draw against your standard workspace usage allowances.

• Describe tasks with clear success criteria and attach background files using the attachment icon.
• Set up recurring scheduled tasks to review calendars, compile feedback, or run routine analyses.
• Track operational status across the In progress, Scheduled, and Completed dashboard views.
• Inspect live execution by opening the dot's cloud computer directly from the desktop profile.

Interface Options and How to Use ChatGPT Dots Remotely

You can communicate with your dot through ChatGPT desktop software, web browsers, mobile applications, Slack, Microsoft Teams, and voice calls.

Context carries across every connected channel, allowing you to assign a project from desktop web and receive milestone notifications in Slack. While your dot can send you messages, progress updates, or decision requests across these platforms, OpenAI notes that dots cannot initiate outbound voice calls to you at launch. When you want to discuss a project verbally, you must start the voice call yourself from the ChatGPT interface.

OpenAI also offers a limited texting beta through a third-party SMS provider for United States Pro subscribers, excluding Business and Enterprise workspaces. You can enable texting by connecting your phone in the desktop application, and message and data rates may apply. OpenAI advises caution when sharing sensitive information by text, and you can reply STOP to stop texts.

Custom Rules, Permissions, and Safety Controls

Custom Rules govern the operational autonomy of your dot by defining whether actions proceed automatically, await pre-approval, require explicit confirmation, or hand off entirely to you.

OpenAI classifies pre-approved actions as instructions you explicitly requested within your prompt. Custom Rules apply to sensitive categories such as sharing data, buying items, or accessing private accounts, and can be adjusted anytime within the mobile app under Customize and Custom rules. Auto-review checks actions that could affect your accounts or share information against your instructions, Custom Rules and safety requirements, ensuring that critical operations like changing account passwords remain strictly in human hands.

During idle periods, dots engage in proactive research to surface helpful project ideas using connected software. OpenAI restricts proactive research tools to read-only access, meaning the agent cannot alter application contents, transmit messages, or manipulate browser windows while conducting background scans. Even with strict policies in place, OpenAI notes that dots can still make mistakes, making human oversight necessary for consequential business decisions.

• Take action without asking: Permits the agent to execute specific steps autonomously.
• Take action if pre-approved: Allows execution only if the action was directly stated in your original prompt.
• Ask before taking action: Halts execution and requests manual confirmation before proceeding.
• Hand off to you: Stops automated execution completely and returns the step to the user.

Local Computer Access and Cloud Isolation Boundaries

Dots operate inside an isolated cloud computer by default, so your computer and its contents stay separate unless you choose to connect it.

If you want your agent to interact with local files or desktop applications, you must enable local access manually through the ChatGPT desktop application. You confirm "Allow access" in the ChatGPT desktop app on that computer, and "Revoke access" stops it. Enabling local access allows the dot to execute local skills, launch Codex workflows, and use your local browser whenever its cloud browser encounters access blocks.

For web authentication, dots can log into supported websites using stored credentials without revealing passwords to the underlying model. OpenAI says safeguards help defend against malicious instructions, and monitoring can pause or stop your dot's work if it detects a safety concern.

Pausing and Resetting Your Dot

You can pause or erase your dot through the profile settings menu whenever you need to suspend operations or wipe stored context.

To pause active tasks, select the three dots menu in your dot profile and click Pause, which halts all background workflows until you tap Paused to resume. Resetting the agent is a permanent action that deletes the dot, wipes past conversations, clears saved memories, and cancels active scheduled tasks. Simply disconnecting a third-party app does not delete the information your dot previously gathered; clearing that historical context requires a full agent reset.

Data privacy controls differ across personal and commercial accounts. OpenAI states that inputs from ChatGPT Business, Enterprise, and Edu workspaces are not used to train models by default. Subscribers on individual Pro tiers can adjust their settings to determine whether conversation data and completed deliverables contribute to model training, and OpenAI says it doesn't train directly on proactive research or your dot's notes to itself.

Account Eligibility and How to Use ChatGPT Dots by Region

Access to dots depends on your subscription tier and geographical location, with personal Pro plans subject to regional restrictions.

OpenAI offers dots to individual users on ChatGPT Pro tiers, which include Pro 100 at $100 per month, Pro 200 at $200 per month, and Pro 500 at $500 per month. However, individual Pro users located in the European Economic Area, Switzerland, and the United Kingdom are excluded from the initial rollout. In contrast, team members with a Premium seat in ChatGPT Business, priced at $125 per user per month, or $100 per user per month billed annually, with a two-user minimum, can access dots across all supported regions without European exclusions.

Enterprise, Edu, and Healthcare organizations can access dots through an administrator-controlled beta toggle that remains turned off by default. Free, Go, and Plus are not listed as including a dot. Subscription rates and tier inclusions change periodically, so check the official ChatGPT pricing page to verify current terms before subscribing. The older dot icons in the ChatGPT interface are unrelated to agent dots, as explained in ChatGPT dot meanings.

Operational Boundaries and Implementation Governance

Evaluating whether an always-on agent fits your workflow requires analyzing task frequency, system integration risks, and monthly operating costs.

At Layer3Labs, we build custom agents and workflow integrations for small and mid-sized businesses, and we consistently see teams struggle when they grant autonomy before testing tool permissions. While OpenAI states that dots can manage cross-application tasks independently, real-world deployment data is not yet available because the feature launched today. Organizations with strict data compliance obligations or teams requiring fixed output guarantees should review security configurations through an AI security audit before connecting primary customer systems.

A dot is not for anyone unwilling to pay at least $100 per month for Pro 100 or $125 per user per month for a Business Premium seat. If OpenAI expands dot availability to entry-level paid tiers or removes regional restrictions on European Pro accounts, our deployment recommendations would expand accordingly. To begin using your agent effectively, create your dot on desktop, set Custom Rules to ask before taking action on sensitive software, and give it one project first.