OpenAI introduced Dots at DevDay on Tuesday, a personal agent that runs continuously in the background and reports back through ChatGPT, Slack and Microsoft Teams.
Each Dot gets its own cloud computer with a browser, terminal and file storage. Powered by GPT-6 Astra, it researches, writes documents, analyzes data and builds software through Codex without waiting for a prompt. OpenAI describes the result as closer to messaging a coworker than operating a tool.
The launch mirrors Meta’s Muse, which arrived three weeks earlier with a similar pitch. Muse also runs on a secured computer, works in the background and asks for approval before sensitive actions. OpenAI leans on the same friendly framing, giving every Dot a name and a cartoon avatar.
Differences show up in the plumbing. Dots reach more than 4,000 apps through plugins, can be joined on a voice call, and keep context across channels and devices. OpenAI is testing specialist Dots with their own identities and credentials for narrow roles such as purchasing, invoice processing and contract management. An integration with Microsoft’s Agent 365 management platform is in the works.
Guardrails rest on Custom Rules that allow, gate or block individual actions, plus an automated review of anything touching accounts or sharing data. Saved passwords can be used without the model seeing them, and a monitoring layer can pause work. OpenAI concedes Dots make mistakes and says results with real consequences should be double-checked.
Pro and Business Premium subscribers get Dots first, with usage limits expanded for the first month.
