OpenAI has introduced Dots, a new class of always-on AI agents designed to perform tasks independently across various applications.
Key facts
- •Dots connect to more than 4,000 apps and maintain context across ChatGPT, Slack, and Microsoft Teams.
- •Each Dot is assigned a unique name and a cartoon avatar to facilitate a coworker-like interaction.
- •OpenAI plans to eventually allow users to deploy multiple agents, though currently, each user is limited to one.
- •Dots are designed to compete with Meta's Muse agent, which launched three weeks prior to the DevDay announcement.
- •The agents can perform tasks such as fixing software bugs, preparing invoices, or creating social media content.
OpenAI unveiled its new 'Dots' AI agents at the company's DevDay 2026 conference. These always-on assistants operate on their own cloud computers to manage tasks, analyze data, and interact with over 4,000 applications in the background. Designed to function similarly to a coworker, Dots can be accessed via voice or text through ChatGPT, Slack, and Microsoft Teams, with the ability to learn user preferences over time.
Capabilities and Operation
Dots are powered by OpenAI's GPT-6 Astra model and feature a dedicated cloud computer equipped with a browser, terminal, and file storage. This allows them to perform research, write documents, and build software independently. Users can monitor the agent's progress through a desktop interface that displays the agent's active work alongside a chat window. While Dots can act autonomously, users retain the ability to intervene at any time by selecting a 'Take over' option.
Safety and Customization
OpenAI has implemented built-in rules for autonomous actions, allowing users to set custom restrictions or define specific scenarios where the agent must request permission. The agents also include an 'auto-review' feature to verify that actions comply with safety requirements. While OpenAI is currently releasing the GPT-6.1 Sol model, it noted that the more advanced GPT-6.1 Astra is being delayed due to safety concerns.
Advertisement
This article was independently rewritten by ManyPress editorial AI from reporting originally published by The Decoder, The Verge AI.
