İçeriğe geç / Skip to content / Zum Inhalt

What Is OpenAI Dots? Open-Source Open Dots Alternatives

Ahmet Balaman

7 min read

Vibe CodingOpenAI DotsOpen DotsAI AgentOpenAI Dev DayOpenClawAI
What Is OpenAI Dots? Open-Source Open Dots Alternatives

At its Dev Day event on September 29, 2026, OpenAI introduced persistent AI agents it calls "dots." Each dot has its own cloud computer and browser and keeps working on your goal even when you are not chatting. Sam Altman described it as being "like delegating to a high-agency engineer or a chief of staff" (CBS News).

This article summarizes the Dots and Open Dots sections of the AI Search channel's weekly AI news video on YouTube and extends them with OpenAI's announcement details and the README pages of two open-source projects; the video link is at the end. This is a news roundup, and I have not tested the product. Information comes from launch roundups and project pages.

What is OpenAI Dots?

With an ordinary chatbot you write a prompt and wait for an answer. With Dots the agent acts like a persistent worker that keeps going while you are away:

  • Its own cloud computer and browser, separate from yours. Connecting your own machine is optional.
  • Connections to more than 4,000 apps through plugins.
  • Reachable through ChatGPT, Slack and Microsoft Teams. Phone calls and texting are coming soon.
  • Powered by GPT-6 Astra.

Example jobs OpenAI gives: watching customer feedback, fixing and testing small bugs and sending a pull request with video documentation; updating launch materials when product scope changes; rerunning scientific analyses as new data arrives; turning interview recordings into short clips and summaries. In the announcement, an early tester had a dot find an unsent invoice, draft it and send it after giving approval.

According to the presenter in the video, this approach is OpenClaw-style agents wrapped by OpenAI into an easy-to-use product. Meta's Muse and Google's Gemini Spark are heading the same way. For a similar open-source terminal agent, see What is Hermes Agent.

Who can use it, and at what price?

  • Access: Pro and Business Premium users in eligible markets. On Enterprise, Edu and Healthcare workspaces a beta is available if an administrator turns it on; it is off by default.
  • Region: The video says "not the EU." According to one roundup, the European Economic Area, Switzerland and the United Kingdom are excluded on Pro, while Business Premium is open in all supported regions.
  • Price: The first dot is included in the plan. Pro plans start at $100 a month. Later you will be able to add more dots and scale each by speed or monthly workload; pricing has not been announced.
  • Usage limit: Conversations with a dot do not count against your ChatGPT usage limits. Tasks a dot starts in Codex or ChatGPT Work do. There is an extended allowance for the first month after launch.
  • Setup: You create your first dot in the ChatGPT desktop app or a desktop browser, connect your apps, then message it from your phone.

Not available yet: texting a dot, multiple dots per user and paid scaling options. For organizations, "specialist dots" (dots with a specific role, their own identity and credentials) are in a pilot stage.

Safety and control: what can an agent do on its own?

Running an agent by itself overnight means spelling out what it may touch. The controls OpenAI announced:

  • Isolation: The dot's cloud computer is separate from yours. On supported sites it can sign in with saved passwords without exposing them to the model.
  • Custom Rules: You can allow specific actions, require approval for them, or block them outright.
  • Activity View: Shows progress, including background work.
  • Auto-review: Actions that could affect your accounts or share information are checked against your instructions and safety requirements; sensitive tasks such as changing a password always stay with you.
  • Monitoring: Built-in security monitoring defends against malicious instructions and can pause or stop a dot's work.
  • Data use: Business, Enterprise and Edu content is not used for model training by default. On personal plans you control whether dot conversations are used.

OpenAI itself states that dots can make mistakes and that consequential work should be reviewed. A practical suggestion: start with read-only access and actions that need approval, and widen permissions as trust grows. I covered agent security in more detail in the Hermes Agent security article; the same logic applies here.

Open Dots: open-source alternatives

Just hours after Dots, two separate open-source projects appeared. Both carry the same name but are independent of each other, and both are early.

1. Open Dots (Anil-matcha). A self-hosted "personal AI agent workspace." According to the README it brings chat, tool use, approvals, connectors and computer tasks into one app. Highlights: a deny-by-default action gateway; approval prompts and audit events for higher-risk actions; app connections through Composio; an optional Docker and Playwright computer runtime; encrypted provider credentials and local SQLite storage. The project calls itself an "early prototype"; multi-user hosting and isolation against a hostile web are not production-ready. It is not affiliated with OpenAI. Requirements: Node.js, Python 3.10+ and a model API key. Project: github.com/Anil-matcha/Open-Dots.

2. OpenDots (CopilotKit). "An open-source template for persistent AI agents, each with its own computer, moving between text, calls and Slack." MIT licensed and in alpha; built with CopilotKit and AG-UI. It includes specialist dots (you define a name, role, instructions and permitted tools), "Space" pages for shared working documents and a computer for each dot. The README says plainly: "a template, not a hosted product"; you run and configure the infrastructure. Project: github.com/CopilotKit/OpenDots.

To avoid mixing the two up, look at the GitHub account: Anil-matcha and CopilotKit. As the video stresses, these are frameworks; you have to connect a language model for them to work.

Which should you choose?

OpenAI Dots Open Dots (Anil-matcha) OpenDots (CopilotKit)
Type Ready-made product Open-source prototype Open-source template (MIT)
Setup Inside ChatGPT You set it up You set it up
Model GPT-6 Astra A model you choose A model you choose
Maturity New, rolling out Prototype Alpha
Cost Included in plan (Pro from $100/month) Software free; model and infrastructure costs are yours Software free; model and infrastructure costs are yours

In short: for a seamless, ready experience, OpenAI Dots; for control, privacy and experimenting, the open-source projects. But it is too early to call either an alternative that is ready for production; the projects' own READMEs say so.

For the same week's model news, see the GPT-6.1 Sol article.

Frequently Asked Questions

What is OpenAI Dots?

A persistent AI agent with its own cloud computer and browser that keeps working when you are away. It runs on GPT-6 Astra and is reachable from ChatGPT, Slack and Teams.

Is Dots free?

No. The first dot is included in the Pro and Business Premium plans; Pro plans start at $100 a month. Prices for extra dots will be announced later.

Who can use Dots?

Pro and Business Premium users in eligible markets. For Enterprise, Edu and Healthcare there is a beta an administrator can turn on.

Is Dots safe?

OpenAI offers isolation, custom rules, approvals and monitoring, but also states that dots can make mistakes and consequential work should be reviewed.

Is there an open-source alternative to Dots?

There are two projects with the same name: Anil-matcha's Open Dots (a prototype) and CopilotKit's OpenDots (an MIT-licensed alpha template). You host both yourself and connect a language model.

Source

Comments