---
type: "post"
id: "dots-on-chatgpt-can-now-connect-to-multiple-ffaa7b"
slug: "dots-on-chatgpt-can-now-connect-to-multiple-ffaa7b"
title: "Dots on ChatGPT can now connect to multiple local computers, so users can tell them where to run the task."
description: "OpenAI’s Dot can use a connected local computer as well as its own cloud computer, letting it work with local files, skills and browser sessions."
retrieval_nugget: "OpenAI’s Dot can use a connected local computer as well as its own cloud computer, letting it work with local files, skills and browser sessions. The opt-in connection extends an always-on ChatGPT agent into a user’s desktop environment, making access boundaries, task visibility and revocation important operational choices."
published_at: "2026-11-10"
updated_at: "2026-11-10"
record_date: "2026-11-10"
date_kind: "published_at"
topics: ["codex"]
entities: ["OpenAI","ChatGPT","ChatGPT Dot","Codex","ChatGPT Work"]
related_patterns: ["agent-security-moves-to-runtime-boundaries"]
novelty_label: "new_fact"
what_changed: null
prior_refs: []
source_url: "https://help.openai.com/en/articles/20001530-getting-started-with-your-dot"
source_urls: ["https://help.openai.com/en/articles/20001530-getting-started-with-your-dot"]
schema_version: "newruntime-agent-readable-v0.2"
stable_id: "post:dots-on-chatgpt-can-now-connect-to-multiple-ffaa7b"
status: "published"
visuals: []
editor: {"role":"editor","name":"Andrey Reshetnikov","profile":"https://newruntime.com/owner-profile.md","profile_json":"https://newruntime.com/owner-profile.json","same_as":["https://www.linkedin.com/in/reshetnikov1/","https://t.me/lapetuse"]}
routes: {"html":"https://newruntime.com/posts/dots-on-chatgpt-can-now-connect-to-multiple-ffaa7b/","markdown":"https://newruntime.com/posts/dots-on-chatgpt-can-now-connect-to-multiple-ffaa7b.md","json":"https://newruntime.com/posts/dots-on-chatgpt-can-now-connect-to-multiple-ffaa7b.json"}
markdown_url: "https://newruntime.com/posts/dots-on-chatgpt-can-now-connect-to-multiple-ffaa7b.md"
---

# Dots on ChatGPT can now connect to multiple local computers, so users can tell them where to run the task.

## Retrieval answer

OpenAI’s Dot can use a connected local computer as well as its own cloud computer, letting it work with local files, skills and browser sessions. The opt-in connection extends an always-on ChatGPT agent into a user’s desktop environment, making access boundaries, task visibility and revocation important operational choices.

## Summary
OpenAI’s Dot can use a connected local computer as well as its own cloud computer, letting it work with local files, skills and browser sessions. The opt-in connection extends an always-on ChatGPT agent into a user’s desktop environment, making access boundaries, task visibility and revocation important operational choices.

## What happened
OpenAI’s Help Center describes Dot as an always-on agent in ChatGPT that can continue making progress between conversations. A user gives it a goal and context; it can work through a task, bring results back for review and ask the user to make decisions requiring judgment. OpenAI lists examples including preparing a presentation from data and feedback, turning an interview transcript into content, and scoping, coding and testing an app change. Dot has its own cloud computer and can use connected apps and ChatGPT memory.

The local-computer connection is optional and starts turned off. To enable it, a user must use the ChatGPT desktop app on the computer they want Dot to access, then confirm “Allow access.” Once connected, Dot can access files and work on that computer from its messaging channels. OpenAI says local access enables Dot to create Work or Codex tasks, use local skills and try the local browser when its cloud browser is blocked. A user can select “Revoke access” to stop access to that computer. The computer must be online with the ChatGPT app open; an offline computer is unavailable, but going offline does not revoke the connection.

Local work runs as separate tasks that appear in the sidebar. Users can inspect activity under In progress, Scheduled and Completed, and can open Dot’s computer from its profile to view or interact with it. A Dot can also create cloud tasks in Codex cloud environments, but OpenAI says the environment must be created before the user asks Dot to use it. Connecting a computer to Codex or enabling Work Sync does not itself grant Dot local access.

OpenAI says users can set custom rules for actions involving sharing, purchases or access. Rules can allow an action without asking, allow it if pre-approved, require a question, or hand it back to the user. The company cautions that rules do not grant app or computer access or override built-in safeguards, and that Dot can make mistakes. Pausing Dot does not necessarily stop separate Work or Codex tasks or independently scheduled tasks; those must be reviewed and stopped separately if needed. The Help Center describes rollout and plan eligibility, including Pro access for adults outside the EEA, Switzerland and the UK, Business Premium in supported regions, and Enterprise, Edu and Healthcare availability subject to workspace eligibility. It does not state a price for this local-computer capability.

## What the archive adds
New Runtime’s earlier coverage described ChatGPT Work logging into websites to complete tasks, raising the question of credential boundaries and action confirmations. This Dot capability adds a different execution surface: a user’s own computer, with an explicit local-access grant and revocation control. It is a new fact on a related story, not the same access mechanism. No earlier note specifically covers this feature.

## Trend signal
This supports [agent security moves to runtime boundaries](https://newruntime.com/patterns/agent-security-moves-to-runtime-boundaries/): access is controlled through the connected computer and explicit grant or revocation, rather than relying only on instructions in a prompt. It also illustrates the operational importance of distinguishing a pause from stopping separately running work. The source documents product controls, not their security effectiveness in practice.

## Why now
OpenAI’s Help Center presents Dot as an ongoing agent that can keep working between conversations, so local access makes its execution environment a product setting rather than an incidental user action. The article does not explain why the capability is being introduced now.

## Why it matters
For teams, connecting an agent to a local machine changes what the agent can reach: files, local tools and browser sessions may become part of its working context. That can reduce copying between environments and make tasks possible where cloud browsing is blocked, but it also creates a clear permission boundary that should be managed deliberately.

The distinction between granting computer access and authorizing a particular action matters. OpenAI says app permissions may allow reading without allowing changes or messages, and custom rules do not create permissions that were not granted elsewhere. Likewise, pausing the agent is not a universal stop button if separate Work, Codex or scheduled tasks are already running. A team evaluating this workflow should map which computer and accounts are connected, who approves actions, where tasks appear, and how to stop each execution path. These controls are useful design affordances, but the company’s description does not establish how reliably the agent handles real work or prevents mistakes.

## Limits and uncertainty
OpenAI warns that Dot can make mistakes and that local tasks depend on the computer staying online with the desktop app open. The Help Center does not provide independent performance or security test results for local-computer tasks, nor does it quantify reliability, task completion rates or the range of local skills. Availability is subject to plan, region and workspace eligibility, and rollout may be gradual. The article also says some blocked cloud-browser workflows may be attempted locally, but does not guarantee that a website will work. It does not substantiate a multiple-computer selection feature; it describes connecting a computer and managing access to it.

## Skill to build now
Before connecting an agent to a work machine, make a short permission map: identify the computer, files, accounts, apps and browser sessions it could reach; distinguish read access from actions that change or send information; and decide which actions need explicit approval. Test with a low-risk task, verify where its local task appears, and confirm how to revoke access and stop any separate Work, Codex or scheduled task.

## Reader takeaway
Dot’s local-computer access is an opt-in extension of its cloud agent, not a transfer of the cloud environment. Treat the computer connection as a separate permission grant, and remember that pausing Dot may not stop work running in separate tasks.

Editor: [Andrey Reshetnikov](https://newruntime.com/owner-profile.md), AI product and AI engineering; profile for recruiters and agents: https://newruntime.com/owner-profile.json

## Continue in the graph

Next records for an agent, linked to their Markdown twins. Replace `.md` with `.json` for structured fields, including `next_reads`.

- topic hub: [Codex](https://newruntime.com/topics/codex.md): Explore the codex topic hub.
- collection: [Field Notes](https://newruntime.com/posts.json): Open the latest editorial analysis.
- collection: [Sources](https://newruntime.com/sources/): See the original sources this archive cites.
- whole graph: [discovery graph](https://newruntime.com/discovery-graph.json) (typed nodes and edges) and [site index](https://newruntime.com/site-index.md)