OpenAI Dots: Always-On Agents, Plans, and API Rates
OpenAI Dots deploy always-on persistent agents on dedicated cloud computers powered by GPT-6 Astra across ChatGPT Pro, Business Premium, and Enterprise plans.
OpenAI announced GPT-6 on October 7, 2026, providing the underlying GPT-6 Astra model that powers Dots, OpenAI's always-on persistent agents. Each dot operates on its own dedicated cloud computer, taking proactive actions across workplace tools instead of waiting for continuous user prompting. Our model pricing data tracks API token prices across providers.12
How always-on Dots run on dedicated cloud computers
A user's dot functions on an isolated virtual machine, operating independently between user sessions. Rather than resetting context after each reply, the agent starts by initializing context directly from user ChatGPT memory. It proactively figures out next steps and continues making progress between conversations.2
Execution relies on Codex alongside external software connections, allowing the agent to write code, call APIs, and manipulate environments directly on its dedicated machine.2

Connecting Dots across web, mobile, Slack, and Microsoft Teams
Users can engage their dot directly inside ChatGPT through chat messaging or voice calls across web, mobile, and desktop interfaces. OpenAI also confirmed that SMS text messaging support for dots will arrive in a subsequent update.2
Workplace collaboration integrations differ by host software. Within Slack channels, dots follow up on questions and pick up next steps. Inside Microsoft Teams channels, dots turn conversations into progress directly within team workspaces.2
Setting execution guardrails with Custom Rules and Auto-review
Operational boundaries are configured through Custom Rules, where administrators and users specify whether certain actions are permitted automatically, require manual approval, or remain completely blocked. Teams evaluating autonomous systems alongside AI agent security risks must weigh these explicit administrative controls before enabling persistent background actions.2
Any proposed action capable of modifying accounts or transmitting data triggers an Auto-review checkpoint. This background check validates execution against user instructions, established Custom Rules, and platform safety criteria. However, sensitive account operations such as modifying passwords cannot be run autonomously by dots and require direct user intervention.2
Eligible plans, administrator enablement, and model pricing
Access to Dots is rolling out across ChatGPT Pro, Business Premium, and Enterprise subscription tiers in eligible markets. While Pro and Business Premium accounts configure dots directly, Enterprise organizations require workspace administrator enablement before individual team members can activate their agents.2
For individual power users, the ChatGPT Pro subscription includes access to Dot as an always-on agent, pro reasoning powered by GPT-6 Astra, and a choice among 3 usage tiers. Teams weighing build costs can compare these against AI marketing agents.3

In our model pricing data, recorded on October 3, 2026, the underlying OpenAI gpt-6-astra API costs $10 per million input tokens and $50 per million output tokens. This places Astra at a premium over mid-tier models like gpt-6.1-sol at $2 input and $10 output, and entry models like gpt-6-luna at $0.1 input and $0.5 output per million tokens.4