Skip to content
NewsProduct launch

OpenAI launches Dots, an AI agent that keeps working after you close the chat

· by Pondero Newsdesk

The short version

OpenAI's new Dots agent runs on its own cloud computer, connects to more than 4,000 apps, and keeps acting on a user's goals in the background between conversations.

OpenAI launches Dots, an AI agent that keeps working after you close the chat

OpenAI launched Dots at its DevDay event on September 29, 2026, a persistent agent that runs on its own cloud computer and keeps pursuing a user's goals after the chat window closes, per OpenAI's announcement. Unlike a Codex task or a standard ChatGPT session, a dot does not end when the user stops typing.

What

Dots is powered by GPT-6 Astra and connects to more than 4,000 apps through OpenAI's plugin ecosystem, per OpenAI. Each dot gets its own cloud computer and browser, works toward a user's goals continuously, and can be reached through ChatGPT, Slack, or Teams, with text messaging coming soon. OpenAI said users "can start with your primary dot, give it a name, and make it your own," and that it envisions "teams of Dots working together" over time. Dots rolled out Tuesday to Pro and Business Premium ChatGPT users in eligible markets and can be launched from Codex or ChatGPT, according to TechCrunch. Enterprise workspaces, including Edu and Healthcare tiers, can turn on a beta once an admin enables it, per OpenAI's announcement; it is off there by default. OpenAI is also piloting "specialist dots" with their own credentials and access for tasks like procurement and invoice processing, and is working with Microsoft to plug them into Agent 365's governance controls.

Dots hands a standing agent access to your apps and your computer

The trade Dots asks users to make is real: connecting apps gives a dot the ability to act on them without a human approving each step. OpenAI built in guardrails to manage that, including a read-only "proactive research" mode for background work, custom rules that force approval on specific actions, and an auto-review system that can pause or stop a dot if it flags a safety concern; certain tasks, like changing a password, always require the user directly, per OpenAI's announcement. TechCrunch's Lucas Ropek noted that "much of this functionality was already possible through Codex and similar agentic harnesses" and that Dots mainly packages existing capabilities into one persistent, branded product rather than introducing a new one. For teams already running agent workflows through Codex or third-party orchestration tools, the practical question is whether consolidating into one standing, always-connected agent is worth trading the finer-grained control of task-scoped runs.

The launch also drew an unplanned rival response. The domain dot.com, owned by Elon Musk's xAI, currently redirects to the download page for its Grok app; domain-registry records show it changed hands in July, months before OpenAI's announcement, which online commentators read as a jab timed to the Dots launch, per TechCrunch.

What to watch next

OpenAI has not said when "teams of Dots working together" moves from stated vision to a shipping feature, or when the enterprise beta opens beyond admin-enabled pilots. Early user reports on how often the "minimal oversight" framing actually triggers an approval prompt will show whether Dots operates as autonomously in practice as OpenAI's launch materials describe.

Sources