OpenAI Dots Bring Always-On GPT-6 Astra Agents to ChatGPT Pro

At DevDay on September 29, 2026, OpenAI announced Dots: always-on agents powered by GPT-6 Astra, each with its own cloud computer, browser, and access to more than 4,000 apps through OpenAI’s plugin ecosystem. Dots are designed to pursue goals over long horizons—preparing invoices, opening pull requests, revising launch materials—while users stay in control through permissions, activity views, and mandatory approvals for sensitive actions.

This is not a incremental feature toggle. It is OpenAI’s consumer-facing answer to the same shift Microsoft and Google are chasing: work that continues when the user closes the tab.

How Dots are supposed to work

OpenAI describes Dots as “frontier intelligence that have your back.” Each dot runs on dedicated cloud infrastructure, learns from feedback, and can operate 24/7. Users interact through ChatGPT on desktop, web, and mobile; OpenAI says Slack and Teams messaging are supported, with texting coming soon.

Key mechanics from the announcement:

• Own cloud computer and browser, inspectable at any time; optional connection to the user’s laptop when permitted.
• Multi-project concurrency without forcing users to juggle separate threads for every task.
• Proactive research when idle: read-only tool use on connected apps—no sending messages or mutating content without an active task.
• Auto-review checks actions against instructions, custom rules, and safety requirements before execution.

Examples in the post span engineering (customer feedback → tested PRs with screen recordings), product launches (scope changes → revised narratives), science (new data → updated figures), enterprise sales (shifting requirements → updated proposals), and content production (transcripts → clips and social drafts).

Rollout and pricing shape

Dots began rolling out on launch day to ChatGPT Pro and Business Premium users in eligible markets. Enterprise users—including Edu and Healthcare—can access a beta when workspace admins enable it.

OpenAI says the first dot is included at no extra cost on Pro and Business Premium, with an allowance for deeper work and extended limits for the first month after launch. Conversations with a dot do not count toward ChatGPT usage limits; work delegated into Codex or ChatGPT Work still consumes those limits as usual.

Safety, privacy, and enterprise specialist dots

Dots inherit ChatGPT protections plus agent-specific safeguards: separate cloud environments, saved-password flows that avoid exposing credentials to the model, monitoring that can pause work, and built-in rules for when approval is required. Password changes and permanent deletions remain user-only actions.

For organizations, OpenAI previews specialist dots with their own identity, credentials, and systems-of-record integrations—starting with focused enterprise pilots in areas like procurement, invoicing, email marketing, support, and contracting.

OpenAI also says it is working with Microsoft to integrate specialist dots with Agent 365 governance and security controls so IT can manage agents through existing Microsoft tooling.

What builders should watch

1. The unit of product is now “agent + computer”

When your competitor ships a VM-backed agent, a stateless chat wrapper feels thin. Even if you are not on OpenAI’s stack, plan for durable state, tool audit logs, and resumable sessions—patterns OpenAI documents separately in its Agents API.

2. Permission UX becomes the bottleneck

Dots lean on custom rules and auto-review because unconstrained tool access does not scale. If you build internal agents, invest in policy-as-configuration: which domains, which write actions, which always require human sign-off.

3. Proactive background work changes trust calculus

Read-only proactive research is OpenAI’s compromise between helpfulness and horror stories about agents emailing your boss. Document exactly what background pollers can see and how users revoke access.

Competitive context

Dots launch in the same window as Microsoft’s Copilot Autopilot private preview and Google’s continued Gemini rollout—each vendor is betting that persistent agents, not single-turn chat, become the default enterprise interface. OpenAI’s differentiation is tight integration with ChatGPT’s consumer scale, Astra-class reasoning, and a plugin catalog counted in thousands of apps.

Developers building on the OpenAI Platform should also study the separate Agents API beta: durable sessions, OpenAI-hosted sandboxes, MCP tools, and subagent delegation for custom products that are not Dots-branded but share the same harness ideas.

Bottom line

Dots move OpenAI from reactive chat to persistent coworkers backed by Astra, a cloud computer, and thousands of integrations—first for high-end consumer and business plans, then through specialist enterprise pilots. The race is no longer “who has the best model on a leaderboard,” but who ships agents people trust to keep working when they walk away from the keyboard.