Guide 37 · Comparison
OpenAI Dots vs Muse: Every 24/7 AI Agent, Compared
OpenAI's Dots joined the always-on agent race on September 29, 2026. What Dots actually does, every rival you can get today, and the honest answer on where Muse fits.
By Abhit Tomar · Updated November 16, 2026 · 13 min read

Short answer
OpenAI Dots (announced September 29, 2026) are always-on agents inside ChatGPT with their own cloud computer and browser, 4,000+ app connections, and approval-gated actions — rolling out to ChatGPT Pro and Business Premium users. The current alternatives: Meta's Muse (free tier, background goals and notifications), Google's Gemini Spark (AI Pro/Ultra), Google's CC family agent (free Labs experiment), and xAI's Grok Bot teammates.
What “always-on” actually means
A normal chatbot waits for you. An always-on agent keeps going after you close the app: it holds ongoing responsibilities, checks connected apps on its own, and brings finished work back for approval. The industry landed on this idea from several directions at once — scheduled tasks, background browser agents, and persistent memory — and in September 2026 it became a product category with a name.
Three things separate a true always-on agent from a chatbot with reminders: it acts without a fresh prompt, it keeps state across days (not just one conversation), and it can touch your other apps through connections rather than asking you to paste things in.
The useful way to think about it is as a spectrum, not a switch. At one end sit reminders and scheduled tasks: the agent does one thing at one time because you told it to. In the middle sit recurring goals: watch a price, follow a story, check a mailbox every morning, and report back only when something changed. At the far end sit persistent agents with standing responsibilities: a project they own, a set of apps they can reach, and the judgment to take the next step without being asked. The further along the spectrum you go, the more the agent needs memory, permissions, and a way to reach you — and the more carefully you should set its limits.
September 2026 is when the far end of that spectrum became a product you could buy. OpenAI shipped Dots, Google had already announced Gemini Spark and expanded its CC family agent, xAI was pitching Grok Bot teammates, and Meta's Muse was running background goals for its users. None of these products is the finished form of the idea — they are the first generation, each with a different answer to the same questions: what should the agent be allowed to do alone, how does it tell you what it did, and what happens when it gets something wrong. This guide compares the answers as they stand, not as their launch videos promise.
OpenAI Dots: the verified facts
Announced at OpenAI's DevDay keynote on September 29, 2026, Dots are persistent agents that live in ChatGPT. Here is what OpenAI and press coverage confirm:
Four pieces make a dot different from a chatbot with good memory, and they are worth understanding before you judge the price. First, a standing goal: you assign ongoing responsibility once, and the dot figures out the next steps itself instead of waiting for a fresh prompt. Second, its own cloud computer and browser, so it keeps working while your laptop is shut — and you can open that computer any time to watch what it is doing. Third, memory that compounds: the dot receives your ChatGPT memories and forms its own as it works, including from connected apps. Fourth, proactive research: when you are not actively working with it, the dot looks for ways to help with read-only access — it can look, but it cannot send messages, edit files, or change your apps until a task and your rules say so.
- Each dot runs on GPT-6 Astra and gets its own cloud computer and browser, so it can research, draft, and code after you've closed your laptop.
- Dots connect to more than 4,000 apps through OpenAI's plugin ecosystem, and can reach you via ChatGPT, Slack, and Microsoft Teams — texting and voice calls are coming.
- You set approval rules: Dots ship with defaults for when to act alone vs. ask, and you can add custom rules to allow, gate, or block specific actions. An auto-review system checks actions against your rules.
- When idle, a dot does “proactive research” with read-only access — it can look but not send messages, edit files, or change your apps without approval.
- Rollout started September 29 for ChatGPT Pro and Business Premium users in eligible markets (Pro tiers now start at $100/month after OpenAI's September 29 restructure into Pro 100/200/500, plus $125/seat Business Premium), one dot per user at first, with more dots later. Enterprise, Edu, and Healthcare workspaces can enable it via an admin.
- OpenAI also launched ChatGPT Space: a shared layer where people, ChatGPT, Codex, and Dots work against the same documents and context.
- The launch demo stumbled — the presenter's dot froze live on stage, which is worth remembering when anyone promises you can “walk away and trust it.”
- The complete picture: What Are ChatGPT Dots? — the full explainer covering setup, every feature, plans and pricing, availability, safety rules, and limitations.
Every 24/7 agent you can get right now
Dots didn't appear in a vacuum. Here is the full current lineup as of September 30, 2026 — what each one is, what it costs, and who it's for.
- OpenAI Dots — the newest. Best for: ChatGPT power users and teams who want agents embedded in Slack/Teams with deep app connections. Catch: the price of entry is Pro or Business Premium.
- Meta Muse — launched earlier in September 2026. Best for: personal life-admin (it handles things like finding subscriptions in your email, booking appointments, comparing insurance quotes). Catch: it asks permission before consequential actions, and background depth varies by task.
- Google Gemini Spark — announced at I/O in May 2026. A 24/7 personal agent living in a Gemini tab with background tasks, recurring schedules, and triggers across Gmail, Calendar, Docs, Sheets, Slides, and Drive. Best for: people living in Google Workspace. Catch: reports say US-only, on AI Pro ($19.99) and up.
- Google CC — expanded September 17, 2026. An agent for families and groups with its own Google account and isolated cloud computer; up to six members share context. Best for: households coordinating over email. Catch: a free Labs experiment, so expect rough edges.
- xAI Grok Bot teammates — xAI's entry in the always-on race, pitched as AI teammates with blob-style avatars. Best for: X-centric workflows. Catch: the least documented of the bunch so far.
- ChatGPT's built-in agent mode — scheduled tasks and a cloud browser inside ChatGPT on lower tiers. Best for: a taste of background agents without the Pro price. Catch: largely superseded by Dots for Pro users.
Where Muse honestly fits
Muse belongs on this list, with honest boundaries. Per Meta's official FAQ, Muse is free with a usage limit, asks permission before sending messages, making purchases, or sharing information — and it keeps working in the background after you close the app. Set a goal (watch a price, follow a story, remind you weekly) and Muse works it in the background, notifying you only when something is meaningfully new.
What Muse doesn't give you: its own always-on cloud computer, autonomous multi-app marathons, or the deep enterprise plumbing (Slack/Teams-native agents, admin governance) that Dots and Gemini Spark are selling. Muse is a personal assistant with background superpowers, not an employee that never sleeps.
The practical read: if your “24/7” need is “keep an eye on this and ping me,” Muse's goals plus notifications cover it free. If your need is “run my team's workflows across forty apps while I'm asleep,” that's the Dots/Spark pitch — at their prices. Try the free thing first; most people overestimate how much autonomy they actually want.
There is also a workflow difference worth naming. A dot is built around projects it owns over days: it keeps a standing goal, works through the steps, and reports back in Slack or Teams where your team already is. Muse is built around your life: goals you set in plain language, notifications when something is meaningfully new, and a chat thread you can pick up any time. The first shape fits a work project with a deadline; the second fits the background chores of a household. Neither is a smaller version of the other — they are answers to different questions, and the honest test is which question you are actually asking.
Which one should you pick?
Match the agent to the job, not the hype:
- Life admin on a budget → Muse (free tier, background goals, approval-gated actions).
- Team workflows in Slack/Teams → Dots, if you're already paying for Pro or Business Premium.
- Google Workspace household → Gemini Spark (AI Pro) or the free CC Labs experiment for families.
- Just curious about background agents → ChatGPT's built-in agent mode or Muse's goals before spending $100/month.
The honest catches
Three things to keep in mind before handing any of these agents the keys. First, autonomy is a dial, not a switch: every serious agent here gates consequential actions behind approvals — if yours doesn't, that's a red flag, not a feature. Second, always-on means always-spending: background work burns tokens or usage budget around the clock, so check what idle agents cost on your plan. Third, the demos are the best day these products will ever have — Dots froze on stage at its own launch. Start every agent with read-only, low-stakes jobs and promote it to real responsibilities only after it earns your trust.
For the Muse-specific side of this — what it can and can't do in the background — see what Muse AI is and the invite-code guide. For the other agents in this race, see what ChatGPT Dots are, what Instinct is, and what Wajo is — plus the full-parameter Muse vs Wajo and Muse vs Instinct comparisons. For breaking developments, watch the updates feed.
How a dot works day to day
Setting one up takes a desktop: you create a dot in the ChatGPT desktop app or desktop web — mobile setup is not supported at launch — give it a name, and it gets a handle based on that name. Then you connect apps deliberately, starting with the few its first goal needs, and set its Custom Rules before giving it real work. For each kind of action you pick one of four behaviors: act without asking, act if you explicitly asked for it in the prompt, ask before acting, or hand it off to you. OpenAI's own example rule is “never send emails,” and the defaults are the safe starting point — tighten them before you loosen them.
In daily use, the dot works in the background and reaches you where you already are: inside ChatGPT, or in Slack and Microsoft Teams, with texting and voice calls promised later. When it is idle it does proactive research with read-only access, surfacing suggestions rather than taking action. You can pause it from its profile at any time, and OpenAI says its safety monitoring can pause or stop a dot's work if it detects a concern. The practical habit is to check in on what it did the way you would check a junior colleague's work: read the summary, spot-check the sources, and only then let the next step run.
What a dot can't do yet
The launch limits are as informative as the features. You get one dot per user at first — teams of dots are an announced intention, not a product. A dot cannot initiate phone calls, and it has no email address of its own; you can connect your personal email account for its tasks, but you cannot give it a standalone inbox. It cannot change your password, and OpenAI says some sensitive steps always stay with the human whatever your rules say.
The money side is the least settled part. The first dot costs nothing extra during the launch window for eligible Pro and Business Premium users, and dot usage does not count against your plan allowance for the first month — but OpenAI has not published what dots cost after that window, or what extra dots or faster work will cost. Free, Go, Plus, and standard Business plans are not eligible at launch, and Pro subscribers in the European Economic Area, Switzerland, and the UK are excluded from the consumer rollout for now. If the price is the question, the honest answer is that it has not been published yet.
Who should pay for a dot — and who shouldn't
Pay for a dot if your work already lives in ChatGPT, Slack, and Teams, you are already paying for Pro or Business Premium, and you have a real project with ongoing steps that nobody on your team has time to own — research that never ends, a document set that needs constant tending, a workflow that crosses four apps. That is the job Dots were built for, and the first dot costs nothing extra during the launch window, so the trial is cheap.
Do not pay for one to find out whether you like background agents. ChatGPT's built-in agent mode on lower tiers and Muse's free goals give you the same shape of experience — a task that runs while you are away and a report when it is done — without the subscription. Start there, give the agent a real job for two weeks, and count how often its work saved you time versus how often you had to redo it. If the answer is “often,” a dot is worth the upgrade; if it is “rarely,” you have learned something useful for free.
| Agent | Maker | Price | Availability | Standout trait |
|---|---|---|---|---|
| Dots | OpenAI | ChatGPT Pro (from $100/mo) / Business Premium ($125/seat) | Rolling out since Sept 29, 2026; one dot per user at first | Own cloud computer + browser; 4,000+ apps; Slack/Teams-native |
| Muse | Meta | Free with usage limit; paid tiers available | Launched Sept 2026; invite/region-gated | Background goals with notifications; approval-gated actions |
| Gemini Spark | AI Pro ($19.99) and AI Ultra | US only, per reports | Deep Workspace integration; recurring schedules and triggers | |
| CC | Free (Labs experiment) | Limited rollout | Family/group agent with its own Google account | |
| Grok Bot | xAI | Not disclosed | Early | AI teammates for X-centric workflows |
Keep reading
02 · ComparisonChatGPT Dots vs Grok Bot vs Muse AI: Which Always-On Agent Fits You?Three always-on agents launched within seven weeks of each other. What each one actually does, what it costs, and which one matches your week.Read guide
06 · ComparisonMuse AI vs ChatGPT vs Claude vs Meta AI: Which Assistant Fits Your Work?Four assistants, four different jobs: who finishes tasks, who writes best, who costs what — and which one you should actually use.Read guide
07 · ComparisonMuse AI vs Claude: Which Fits Your Workflow?Choose based on the work you repeat, not a generic ranking.Read guide