OpenAI is turning ChatGPT into a continuously active work platform with Dots, expanded plug-ins, and shared workspaces. The changes matter most if you want to combine research, documents, communication, and recurring tasks in one service. However, the DevDay 2026 announcements range from immediately accessible features to products restricted by subscription or region.
Why Dots are more than new chat characters
Dots are personal agents powered by Artificial Intelligence (AI), with an agent being a system that pursues a task across multiple steps. According to OpenAI, each Dot receives its own cloud computer and browser and can keep working after you log off. The initial report on Dots identifies GPT-6 Astra as the underlying model and says the agents can connect to more than 4,000 applications through ChatGPT plug-ins.
The main difference from a conventional chatbot is persistence. Instead of asking a question and waiting for one response, you define a goal that the Dot can continue pursuing in the background. OpenAI describes Dots as largely independent of any particular device or interface; you can contact them through ChatGPT, Slack, and Microsoft Teams, with text-message support expected later.
OpenAI’s examples show the intended direction. One Dot could monitor customer feedback and work on reported software bugs, while a scientist could have another repeat an analysis whenever new experimental data arrives. A separate report offers a more everyday example: identifying a forgotten invoice and preparing it to be sent.
The underlying idea is not entirely new. TechCrunch describes Dots as a new package for existing agent capabilities, some of which were already possible with Codex. Meta’s recently released Muse agent also runs on a secured computer, continues working in the background, and requests approval for sensitive actions, according to The Decoder.
What plug-ins are meant to add to ChatGPT
Plug-ins are extensions that connect ChatGPT to other services and their data. Existing connections reportedly include Slack, SharePoint, Airtable, and Google Drive. The new extensions can receive a dedicated place in the sidebar and provide interactive panels, allowing you to operate a connected tool without leaving the conversation.
This is intended to make ChatGPT behave less like a text box and more like an interface for multiple applications. Plug-ins can include viewers for their own file formats, and you can approve each extension’s requested access separately. OpenAI also plans to improve discovery and recommendations in the directory, while the expanded plug-in system can also be embedded in lightweight sites created with ChatGPT.
Support for MCP Events is particularly relevant to recurring work. This proposed specification allows a plug-in to respond to an event in a connected application and initiate an automated workflow. For example, a team could instruct a page to check a communication channel and add newly discovered information to a shared document.
The Decoder also reports Team Tasks, an enterprise marketplace with 32 partners, and a “Sign in with ChatGPT” option for using allowances with third-party providers. According to that report, the plug-in changes will be available across all ChatGPT plans. The sources do not say which individual plug-ins will be offered in Switzerland, what languages they will support, or how connected providers will process user data.
What the new work features bring together
Space is a shared workspace inside ChatGPT. Pages and files live together there while colleagues work with ChatGPT and their own Dots. Unlike a conventional online drive, agents are supposed to monitor material continuously and update it according to defined instructions.
Pages is the accompanying document tool for collaboration between people and agents. OpenAI says you can use it to write, research, generate charts, create images, and visualize information. Slides adds presentations that people and agents can create, comment on, and edit together; according to the report on OpenAI’s office tools, Slides was not yet available on September 29 and was due to roll out in the following weeks.
This combination puts OpenAI in more direct competition with Google Workspace, Microsoft Office, Notion, and Slack, even though Microsoft remains a close partner. It represents more than a collection of extra buttons: chat, files, documents, connected applications, and task-performing agents are meant to coexist in one environment. The announcements do not independently establish whether this bundle works better in daily use than specialized tools.
A similar direction is visible in Codex, OpenAI’s agent for software work. Codex is getting reusable cloud environments accessible across devices, with shared approved settings and automated reviews that can continue when a laptop is closed. Codex is peripheral for most general users, but the update reflects the same pattern: tasks and their working environments persist in the cloud instead of starting over with every session.
Pros and Cons of a persistent work platform
Pros:
- Continuity – Dots can continue pursuing a task while you are offline or switching devices.
- Fewer context switches – Documents, presentations, chat, and connected applications are intended to share one interface.
- Automation – Events in connected services can trigger updates and other recurring workflows.
- Collaboration – OpenAI says people and agents can work together on pages, files, and slides.
Cons:
- Broad access – A continuously active agent needs data, tools, and permissions that must be tightly limited.
- Persistent mistakes – A misunderstood goal can produce a chain of incorrect actions rather than one bad answer.
- Dependence – Combining communication, files, and workflows in ChatGPT ties a larger share of your work to OpenAI.
- Unclear availability – Access varies by plan, market, and rollout schedule, while the sources provide no specific prices.
OpenAI says sensitive actions will require fixed approvals, while plug-in access can be authorized individually. Dots may use read access during idle periods to look for useful tasks, which is both practical and uncomfortable: an agent needs sufficient visibility before it can recognize relevant connections. The available reports do not specify where data will be stored, how long it will be retained, or how it will be processed.
What you can use right away
If you are a beginner, the expanded plug-ins are the most sensible starting point, provided they appear in your account. Connect only one service at first, inspect its requested permissions, and choose a limited task such as summarizing approved files or updating a shared page. This lets you evaluate the benefit without immediately handing all communication, documents, and automation to ChatGPT.
If you are an advanced user, you can test a workflow that combines Space, Pages, and one selected plug-in. A process with clear oversight is preferable, such as monitoring a team channel and adding confirmed information to a document. Persistent agents should initially receive read-only access or prepare drafts, with final changes and external actions still requiring explicit approval.
Immediate access to Dots is less clear. One source describes Dots as a managed product that can be deployed now, while another says they are launching for Pro and Business Premium users in eligible markets. The Decoder reports a different limitation: one Dot is included in a regular subscription, but European Pro users initially receive no access and enterprise customers get different terms.
For Switzerland, it therefore remains unclear whether OpenAI will treat the country as part of the stated European restriction; none of the sources provides a separate Swiss policy. They also offer no concrete prices, confirmed German-language availability, or detailed privacy terms. A plan named “Pro 500” is mentioned, but its name alone does not establish a reliable price.
OpenAI’s direction is clear: ChatGPT is meant to store, monitor, and continue work rather than simply answer requests. Plug-ins and shared documents provide tangible steps in that direction, while Dots represent the much larger move toward persistent agents. The potential time savings come with an equally persistent risk: a system with long-term access can carry mistakes, excessive permissions, and unwanted actions far beyond a single chat session.
Sources
- OpenAI Launches dots: Always-On GPT-6 Astra Agents That Work From Their Own Cloud Computers – Medium unknown, 2026-09-29
- OpenAI launches Dots, its bubbly agentic avatar – Medium unknown, 2026-09-29
- OpenAI expands ChatGPT’s plug-ins with app-like interfaces and automations – Medium unknown, 2026-09-29
- Weniger Chatbot, mehr OS: OpenAI baut ChatGPT zur Arbeitsplattform um – Medium unknown, 2026-09-29
- OpenAI takes on Microsoft with the launch of what feels a whole lot like ChatGPT’s own office suite – Medium unknown, 2026-09-29
- OpenAI gives Codex reusable cloud environments that work across devices – Medium unknown, 2026-09-29
- OpenAI schickt Dauer-Agenten “Dots” gegen Metas Muse ins Rennen – Medium unknown, 2026-09-29
Image: Anna Shvets via Pexels