OpenAI Dots: How Always-On AI Agents Work and Who Can Use Them

OpenAI introduced Dots, a new class of always-on AI agents, at its DevDay event on 29 September 2026. Unlike conventional chat assistants that respond to individual prompts, Dots are designed to manage ongoing tasks, remember relevant context, and continue working between conversations.

Powered by GPT-6 Astra, each Dot has its own cloud computer and browser, allowing it to carry out multi-step work across connected applications. OpenAI says Dots can connect to more than 4,000 apps through its plugin ecosystem and interact with users through ChatGPT, Slack and Microsoft Teams. The launch was part of more than 20 announcements spanning ChatGPT, Codex and OpenAI's developer tools.

What OpenAI Dots can actually do

OpenAI describes several examples of how Dots can handle work across applications. In software development, a Dot can examine customer feedback, investigate recurring problems, write and test fixes, and prepare pull requests for human review. For content creators, it can turn interview transcripts into show notes and draft social media posts.

The company also describes research and sales workflows. A Dot can update an analysis when new data becomes available or help revise a proposal as customer requirements change. OpenAI says Dots can work across several projects and bring results back for review, rather than requiring users to direct each individual step.

These are examples of intended capabilities, not independently verified performance guarantees. How reliably a Dot completes a task will depend on its available permissions, connected applications, and the work involved.

Who gets access, and how

According to OpenAI's getting-started documentation, the first Dot is included at no additional cost for eligible ChatGPT Pro and Business Premium subscribers. Pro availability initially excludes the European Economic Area, Switzerland, and the UK. Enterprise, Edu and Healthcare workspaces can try the beta when administrators enable it.

Users can create a Dot through ChatGPT's desktop app, desktop website or supported mobile apps. They can then assign tasks, review ongoing work, and inspect the Dot's cloud computer. Access to a user's local computer is optional and disabled by default. OpenAI also says deeper work uses the relevant subscription allowance, so an included Dot does not necessarily mean unlimited autonomous task execution.

What Dots can access and when approval is required

OpenAI says Dots use the permissions granted through connected plugins. They can also review permitted information proactively, but their background research tools cannot directly send messages, change content through plugins, or control a computer.

For tasks involving external actions, OpenAI applies approval rules. Users can set additional restrictions, such as preventing a Dot from sending emails, although these rules cannot override the platform's built-in safeguards. Sensitive operations may require explicit approval or direct user control.

There are also privacy considerations. OpenAI says disconnecting a plugin stops future access through that connection but does not automatically remove information already incorporated into the Dot's context. Users can pause their Dot and manage connected services, while data use depends partly on their ChatGPT plan and settings.

These controls define important boundaries, but they do not eliminate the possibility of agent errors or unintended actions.

A new $500-a-month Pro tier

DevDay also introduced Pro 500, a $500-per-month ChatGPT plan offering access to Ultrafast, OpenAI's premium speed tier. The company reports up to eight times faster token generation in Codex and up to six times faster generation through its API.

OpenAI says GPT-6 Astra Ultrafast is available through the API and in ChatGPT Work and Codex on eligible Pro 500 and Enterprise plans. The performance figures are company-reported maximums, and actual results may vary by workload. The announcement suggests that faster AI execution is becoming a separate commercial offering alongside improvements in model capabilities.

GPT-6.1 Sol and the delayed Astra update

OpenAI also introduced GPT-6.1 Sol, which it describes as approaching GPT-6 Astra's performance in agentic coding, computer use, and professional work. The company says Sol's standard input and output token prices are approximately one-fifth of Astra's, potentially making it a lower-cost option for some AI workloads.

The announcement followed a Bloomberg report that OpenAI had delayed a planned Astra update over security concerns. OpenAI has not publicly detailed those concerns in its DevDay recap. The reported delay should therefore remain separate from the company's confirmed product announcements.

Why it matters

Dots represents a shift from AI assistants that respond to individual requests towards systems designed to manage ongoing responsibilities. By combining persistent context, connected applications, and computer use, OpenAI is attempting to make AI useful across entire workflows rather than isolated tasks.

For businesses, the potential benefit is continuity. An agent could follow a project as requirements change, update supporting documents and identify work requiring human attention. However, the value depends on reliability, permission management, and the quality of its integrations. Persistent access also increases the importance of controlling what an agent can read, change or share.

The introduction of Ultrafast and GPT-6.1 Sol adds a separate dimension to OpenAI's strategy: giving developers different options for balancing model capability, execution speed and cost.

How Dots fits into a crowded agent market

OpenAI is entering a market where AI companies are increasingly competing on tool use, integrations, and the ability to complete multi-step tasks. Anthropic's Claude ecosystem, Google's agent tooling, and products such as Manus illustrate the range of approaches to AI-driven workflows.

Dots differentiates itself through its persistent working model, dedicated cloud computer and integration with ChatGPT and workplace communication tools. Whether those features provide a lasting advantage will depend on how reliably the product completes real tasks, not on how many applications it can connect to.

What to watch next

Several questions will determine whether Dots becomes useful beyond its early rollout. The first is reliability: how often can an agent complete multi-step work without repeated corrections? The second is control, particularly whether organisations can manage permissions and review actions across connected applications. The third is cost, including how subscription allowances affect sustained autonomous work.

Broader availability, independent performance assessments, and clearer evidence of business adoption will help establish whether always-on agents offer meaningful advantages over conventional AI assistants.