OpenAI Launches Always-On Agent 'dots' Based on GPT-6 Astra

OpenAI has launched dots, an always-on agent based on GPT-6 Astra. dots features its own cloud computer and browser, performing tasks on behalf of users 24/7, and connects to over 4,000 apps through the OpenAI plugin ecosystem.

Key Features and Use Cases

dots is accessible via various channels including ChatGPT, text messages, Slack, and Teams, maintaining context across all channels. It learns user preferences and standards, performing 'proactive research' in the background to handle tasks before they are even requested. During this process, write permissions for apps are restricted to ensure security.

Real-world use cases include developers incorporating customer feedback and generating PRs, responding to scope changes in product launch materials, scientists updating data analyses, reflecting requirements in sales proposals, and automating content creation.

Security and Permission Management

Each dot has a unique ID and cloud computer, while the user's local computer is isolated by default. Users can directly set access permissions and decide whether to allow, request approval for, or block specific tasks through custom rules. Sensitive actions, such as changing passwords, must always be performed directly by the user.

Availability and Pricing

It will be rolled out sequentially to users on Pro, Business Premium, and Enterprise plans in supported regions. The first dot is included in the plan at no additional cost, with usage limits expanded for the first month after launch. Conversations with dots are not included in ChatGPT usage limits, but Codex or ChatGPT Work tasks are included in existing limits. Additionally, enterprise 'specialized dots' will integrate with Microsoft Agent 365 to support enterprise governance and security control features.