OpenAI Dots: What Always-On ChatGPT Agents Are and What People Build With Them

What OpenAI Dots are

Dots are persistent agents that live inside ChatGPT. Press coverage of the launch describes each dot as running on OpenAI's GPT-6 Astra model with its own cloud computer and browser, working toward goals you give it around the clock rather than only when you type a prompt (BGR). The official launch post is at openai.com, though it was not readable to automated fetching, so the details here come from press and community sources.

In OpenAI's own words on X, a dot "learns how you work, keeps context across apps, coordinates Codex tasks, and flags what needs your attention" (OpenAI Developers). Coverage also lists access through desktop and mobile apps, the web, Slack, Microsoft Teams and voice, along with 4,000+ app integrations through plugins and a Custom Rules system where you decide which actions need approval, are allowed, or are blocked (Ecosistema Startup).

Who makes it and when it arrived

OpenAI announced dots at its DevDay keynote on 29 September 2026 (BGR). The launch was accompanied by an official promo video, and by Space, an agent-native office suite inside ChatGPT where people and dots share files (Dan Shipper).

Access and pricing

Reports say dots are available on ChatGPT's Pro and Business Premium plans, with the first dot included and chatting with it not counting toward normal usage limits; Enterprise, Edu and Healthcare workspaces can reach them after administrator approval. They are not on Free, Go or Plus. Press disagrees on the exact Pro tier and price, so check OpenAI's pricing page before buying. Price for additional dots had not been announced at launch (Superframeworks). One review notes the Pro plan excludes users in the EEA, Switzerland and the UK.

What people are building

Within days of launch, public examples fell into a few patterns.

Developer workflows. The most common use is triage and delegation to Codex. One guide wires a Slack bug channel to a dot that hands fixes to cloud Codex, with rules and AGENTS.md constraints set first and a human merging (Codex Workshop). Another proposes a framework for turning bug reports into reviewable GitHub pull requests, insisting that a patch be tied to the exact tested revision (Wavect).

MCP bridges and connectors. Developers are extending dots through MCP. Examples include a handoff bridge across Codex, dots and Grok Bots, a QQ messaging bridge, and an unofficial MCP server that searches dots guides and news. One project even gives a dot a body by connecting it to a Stack-chan desktop robot.

Dashboards and open alternatives. Third parties are building a dashboard for a dot's tasks and output files, and open-source alternatives such as the open-dot Mac app and Poka-Bot. Curated lists like awesome-openai-dots collect workflows and rules.

Personal assistants. Early users describe a single main conversation while the dot coordinates other threads (Angel), and one user praised a named dot as proactive and fast (post). A lighter moment: a dot asked OpenAI's help bot about its own features.

Limits and caveats

• Early reliability is a concern. Dan Shipper of Every said the version he tested was too buggy to recommend at launch, citing permission problems and dropped messages, according to a founder-focused review that advises keeping dots out of critical paths for now. Treat that as a launch-week snapshot.
• Safety guardrails are deliberate: dots reportedly cannot send messages, edit content or control a computer without permission, and sensitive tasks like password changes are left to the user (Ecosistema Startup).
• Regional availability is limited, and prices for extra dots were unannounced.
• Many community tools are unofficial, brand new and have few stars. Be wary of repos offering a "download" of dots; the real product runs inside ChatGPT, not as a Windows installer.
• Our figures come from press and community posts, not the official page, and may change.

Frequently asked questions

What is an OpenAI dot?

An always-on agent inside ChatGPT with its own cloud computer and browser that keeps working on your goals between conversations.

When were dots released?

They were announced at OpenAI DevDay on 29 September 2026 and began rolling out to eligible subscribers.

Who can use dots?

Pro and Business Premium subscribers get the first dot, and some enterprise workspaces after admin approval. Free, Go and Plus are excluded, and the Pro plan is reportedly unavailable in the EEA, Switzerland and the UK.

Can dots work with Codex?

Yes. OpenAI says dots coordinate Codex tasks, and community guides use that to turn bug reports into pull requests with human review.

Are dots reliable yet?

Early testers reported bugs, so keep them away from critical or customer-facing work until they mature.