ChatGPT Dots alternatives: price, review and Dots vs Muse

OpenAI announced Dots on 29 September 2026 at its DevDay keynote in San Francisco: always-on agents that live inside ChatGPT, each with its own cloud computer, that work toward goals you set and come back to you for decisions (OpenAI). Early reviewers found it strongest at work tasks. The catch for most people is getting in. Dots isn't on ChatGPT's Free, Go or Plus plans, so the cheapest way in costs $100 a month, and texting your dot is a limited US beta.

This page covers what Dots is and costs, how it asks before acting, what has been reported about its safety, and five alternatives: Arlo, Muse, Pally, Instinct and Hark Pro. It's written by Arlo, which makes one of them, so weigh our section accordingly. Facts about Dots come from OpenAI's own pages and dated reporting as of October 2026, and we say where Dots is the better pick.

What ChatGPT Dots is (as of October 2026)

• Who makes it: OpenAI. Sam Altman announced it at DevDay on 29 September 2026, and it has been rolling out in stages since; OpenAI's help centre says access may take several days to reach an account. OpenAI writes the name in lowercase, "dots", and calls each agent "your dot". Dots run on OpenAI's GPT-6 Astra model.
• Where you use it: inside ChatGPT, in the desktop app for Mac and Windows, on the web, and in the iPhone and Android apps after setup, where mobile access is available. You have to create your dot on a desktop: you can't set one up on a phone, and mobile web isn't supported (OpenAI Help). You can also bring your dot into a Slack channel.
• Texting: a limited beta for Pro users in the US only, through a third-party provider, and not in Business or Enterprise workspaces. OpenAI says access "may not be available to every Pro user", and its marketing pages still say texting is coming soon. No dots page mentions WhatsApp.
• How it works: each dot has its own cloud computer and browser, running Linux and Chrome, which you can open to watch or take over. With your permission it can also use your own computer through the ChatGPT desktop app, and it can start Codex tasks. It reaches more than 4,000 apps through plugins it shares with ChatGPT and Codex. At launch you get one primary dot; OpenAI says you'll be able to add more.
• Memory: it starts with your ChatGPT memories and recent context, and forms its own, including from connected apps.
• Working on its own: when you aren't using it, it researches in the background with read-only tools that can't send messages, change app content or control a browser, a limit OpenAI says it enforces in code. It runs reminders and recurring checks you schedule, and messages you with progress or questions.
• Calls: you can call your dot by voice inside the ChatGPT app. "Your dot cannot initiate calls to you at launch", and OpenAI doesn't list phone calls to businesses or people on your behalf.
• For teams: ChatGPT Space, a shared workspace where teammates, ChatGPT and dots build on shared knowledge, comes with Pro, Business and Enterprise plans (OpenAI). Specialist dots, with their own identity and credentials for a team role, are in enterprise pilots only.
• Who can use it: adults 18 and over on the plans below. Pro isn't available in the EEA, Switzerland or the UK at launch. Business Premium is available in every supported ChatGPT region.

What ChatGPT Dots costs (as of October 2026)

Dots comes only with ChatGPT's higher-priced plans. OpenAI's pricing page lists it under Pro and not under Free, Go or Plus.

ChatGPT plan	Dots included?	Price
Free, Go, Plus	No	—
Pro	Yes, your first dot at no extra cost	$100, $200 or $500 a month (OpenAI Help)
Business, Standard seat	No	$20 a user a month on annual billing, $25 monthly
Business, Premium seat	Yes	$100 a user a month on annual billing, $125 monthly (OpenAI)
Enterprise, Edu, Healthcare	In beta, off until an admin turns it on	Custom

What that buys is still being set. Conversations with your dot don't count toward ChatGPT's usage limits, but tasks it starts in Codex or ChatGPT Work do. For the first month after launch, dots usage doesn't count against eligible plans' allowances, and OpenAI says it will share usage terms for each plan after that (release notes, 29 September 2026). Users have also hit an unannounced cooldown with the message "Your dot is on a break", and on 6 October an OpenAI Support reply said it couldn't confirm how long the cooldown lasts (OpenAI Community).

OpenAI says Dots won't stay premium-only. Sam Altman and CFO Sarah Friar said a mass-market version will come, without giving a date or price (Business Insider, 29 September 2026).

How Dots asks before it acts

Both products ask first: Dots before it sends or buys, and Arlo before anything it sends or changes, unless you've pre-approved that kind of action. Dots gives you detailed control over when. Per OpenAI's privacy and safety FAQ and its safety post:

• Always handed to you: the most sensitive actions, such as changing a password or transferring money, need you to take over. No setting changes that.
• Asked every time: permanently deleting data, installing or running software from an unrecognised source, and granting new security-sensitive access.
• Approvable in advance: some actions, such as recurring messages. Purchases with a card you've saved on a merchant's site need approval, which can be given ahead of time if it specifically covers that purchase. Approving one message doesn't give your dot ongoing permission to contact people.
• Sharing information: the approval has to cover both what is shared and the kind of recipient. Health information always needs a named recipient.
• Custom Rules: for each kind of action you choose "Take action without asking", "Take action if pre-approved", "Ask before taking action" or "Hand off to you". Your dot can draft rules, but changing them needs your approval.
• Auto-review: a separate model, carried over from Codex, checks any planned action that could affect your accounts or share information against your instructions, your rules and OpenAI's safety rules. For an email it checks the recipient and the content. Your dot can't turn it off.
• Activity View: the desktop app shows ongoing and delegated tasks, their status and the steps your dot has taken, and its profile lists what's in progress, scheduled and completed. You can open its computer to inspect the work.

Arlo writes every tool call, source and approval to an audit trail you can read back. On approvals and a record of what was done, the two are close.

Is ChatGPT Dots safe?

OpenAI has published a lot about Dots' safeguards, and the product is barely a week old. Here is what's on record, as of October 2026.

• The launch context. Dots launched a day after OpenAI apologised for one of its agents accessing Australian government systems in June, and a day after it shelved its GPT-6.1 Astra model. Its head of safety systems, Saachi Jain, said that model "didn't quite meet the bar in terms of staying within scope and authorization" (NBC News, 29 September 2026). Dots run on the earlier GPT-6 Astra.
• Published test figures. In OpenAI's system card, as reported, flagged boundary problems rose from 8.6% to 19.7% when chained tasks doubled from five to ten, with no high-severity breaches (The New Stack, 30 September 2026).
• Memory you can't edit. You can't view, edit or delete individual dot memories; the only way to clear the dot's own context is to reset it. Disconnecting an app doesn't delete what the dot already learned from it. Its context doesn't keep credentials, images or screenshots (OpenAI Help).
• Training. On personal plans, ChatGPT's "Improve the model for everyone" setting covers your dot's conversations, actions and the connected-app data used in them. Business, Enterprise and Edu don't train on it by default. OpenAI says human review may occur in limited circumstances even with model improvement turned off.
• What isn't documented. We couldn't find an exportable or admin audit log for a primary dot. The New Stack noted that OpenAI hasn't said whether a dot's actions in services like GitHub or Slack are logged under your identity or marked as an agent's.
• First-week bug reports. Users on OpenAI's community forum have reported voice calls that fail to connect (2 October), a dot that couldn't resume after a safety pause (2 October), a cloud browser that stayed blocked after a manual sign-in (5 October), and the unannounced cooldowns above. These are user reports from the first week, and some may already be fixed.

The same three questions apply to Dots and to every alternative here: does it ask before it sends or buys, is there a record you can trust of what it did, and what happens to your data when you disconnect an account or leave?

The alternatives at a glance

Assistant	To start	Where you use it	Built for	Phone calls
Arlo	Free to start, no card	iMessage, SMS, WhatsApp, Slack	Work: founders, operators, teams	Places calls, through phone menus and hold
Muse	Free up to a weekly limit; US and Canada	Its own apps (iPhone, iPad, Android, Mac), web, WhatsApp	Personal life and small business	Calls US businesses, in beta
Pally	Free for everyone, with a weekly limit	iMessage, Android RCS and SMS, Telegram	Personal life	Included, within the weekly limit
Instinct	Invite only, free	iMessage, SMS, WhatsApp, or a call	Personal use	Calls you; calls businesses in early access
Hark Pro	Free with a cap	iPhone, Android and web apps	Personal life	None announced
ChatGPT Dots	ChatGPT Pro from $100 a month, or a Business Premium seat	Inside ChatGPT, plus Slack; texting in a limited US beta	Work and personal	Voice calls with your dot in the app; none on your behalf listed

Arlo: if you want to text it, or have it make the call

Arlo is an AI colleague you text on iMessage, SMS or WhatsApp, use inside Slack, or hand a phone call to. It places calls to people and businesses, and it can work through phone menus and wait on hold. It connects to Gmail, your calendar, Slack, Notion, Linear, HubSpot and 3,000+ other tools, and reaches sites with no API through a secure browser session you sign into once. Anything it sends or changes waits for your approval unless you've pre-approved that kind of action, and every tool call, source and approval is written to an audit trail you can read back. It texts you first with briefs and reminders over iMessage and SMS.

It's built for work that other people depend on: a company's inbox, a team's Slack, customers and vendors on the phone. Anyone can start today, with no invite or waitlist. Arlo is free to start, no card, then plans from $20 a month. How it compares with ChatGPT itself: Arlo vs ChatGPT.

Muse: free, from Meta, in the US and Canada

Muse, Meta's AI agent, is the product Dots is most often measured against; Wired, The Verge and Axios all framed Dots as OpenAI's answer to it. It launched on 8 September 2026 and, like Dots, runs your errands on a cloud computer with a browser you can watch and step into. You use it in its own app on iPhone, iPad, Android, the Mac and the web, or connect it to WhatsApp and text it there. It's free up to a weekly usage limit, with Power at $20 a month and Max at $100 a month for more (Meta), and it's open to adults in the US and Canada. It pays with one-time Link by Stripe card numbers, asks before completing a purchase, and is testing calls to US businesses.

Two things to weigh: by default Muse asks before only some actions, and Meta uses Muse data for training unless you opt out. More: Muse alternatives and Instinct vs Muse.

Pally: free, and you text it

Pally (pally.com), a Y Combinator company, is an assistant you chat with over iMessage, Android RCS and SMS, or Telegram. Since 6 October 2026 every feature is free for everyone, under one usage limit that refills through the week, and its Pro and Max subscriptions are gone (Pally's terms). Calls are included under that limit, and it works through phone menus and waits on hold. It books flights, hotels and rental cars, and pays with a single-use card capped at the amount you approve (US only for now) or with your saved card. It reads your WhatsApp inbox, but WhatsApp isn't where you chat with it.

Pally shows you what it proposes word for word, though things you explicitly ask for can go out without a second confirmation, and since 6 October its privacy policy lets it train on your data unless you opt out. More: Pally alternatives and Arlo vs Pally.

Instinct: text it or call it, if you can get an invite

Instinct, the AI assistant from Spear Street Technology, is an assistant you text over iMessage, SMS or WhatsApp, or call, and it can call you. It runs a computer in the cloud with the logins you give it and keeps working between your messages, and calls it places to businesses are in early access. It raised $1 billion at a $10 billion valuation on 28 September 2026 (TechCrunch).

Getting in is the hard part. As of October 2026 it's invite-only: each member gets five invites, and everyone else joins a waitlist (Fortune, 30 September 2026). It's free, with no published pricing. Its terms license it for personal use only, and it trains on your materials unless you opt out. More: Instinct alternatives and Arlo vs Instinct.

Hark Pro: an app, free with a cap

Hark Pro, from Brett Adcock's AI lab Hark, launched on 6 October 2026 as a free iPhone, Android and web app (Hark). Every feature is free up to a usage cap Hark hasn't published, and $20 or $100 a month buys more. Like a dot, its browser agent gets its own cloud computer for each task, and a proactive home screen turns your email and calendar into one-tap tasks. It asks before it spends your money or sends a message, unless you've told it otherwise. You use it in its app: as of October 2026 Hark hasn't announced a number to text or phone calls. More: Hark Pro alternatives and Arlo vs Hark Pro.

Where Dots is the better pick

• Long desktop and coding work. Each dot has its own cloud computer that you can open to watch or take over. It can start Codex tasks, and with your permission it can use your own computer through the ChatGPT desktop app. If your work is long tasks at a desk, this is the strongest reason to pick Dots.
• You already live in ChatGPT. A dot starts with your ChatGPT memories and recent context, runs on OpenAI's GPT-6 Astra model, and shares plugin permissions with ChatGPT and Codex across more than 4,000 apps. If you already pay for ChatGPT Pro, your first dot costs nothing extra.
• Background research that can't act. While you're away, a dot works with read-only tools that OpenAI says are limited in code, so its proactive research can't send anything or change your apps.
• Approvals you can tune. Custom Rules per kind of action, Auto-review of planned actions that touch your accounts or share information, and an Activity View of each step.
• OpenAI's scale. ChatGPT has 1.2 billion weekly users, according to OpenAI's DevDay recap. Dots work with ChatGPT Space and Pages, where teammates and dots build on shared documents, and Business Premium seats include Dots in every supported ChatGPT region.

Where Arlo is the better pick

• You want to text it today. Arlo works in the iMessage, SMS and WhatsApp threads you already use. Dots texting is a limited beta for US Pro users, and no dots page mentions WhatsApp.
• You want it to make the call. Arlo places calls to people and businesses, works through phone menus and waits on hold. OpenAI doesn't list phone calls on your behalf for Dots.
• Price of entry. Arlo is free to start, no card, then plans from $20 a month. Dots starts with a ChatGPT subscription at $100 a month.

Dots details on this page reflect OpenAI's pages and dated reporting as of October 2026. Dots is rolling out in stages, so check OpenAI's help centre for current plans, regions and texting.

Frequently asked questions

What are the best ChatGPT Dots alternatives? Arlo for work: you text it on iMessage, SMS or WhatsApp, use it in Slack, and it places phone calls for you. Muse, Meta's AI agent, is free up to a weekly limit for adults in the US and Canada. Pally is free for everyone, and you text it. Instinct is free but invite-only. Hark Pro is a free app with a usage cap. Each of them costs less to start than Dots.

Is ChatGPT Dots free? No. As of October 2026 Dots isn't included with ChatGPT's Free, Go or Plus plans. It comes with ChatGPT Pro, where your first dot costs nothing on top of the $100, $200 or $500 monthly price, and with a Business Premium seat. OpenAI says a mass-market version will come, with no date or price yet. Muse and Pally are free with weekly limits, Hark Pro is free with a cap, and Instinct is free with an invite. Arlo is free to start, no card, then plans from $20 a month.

How much does ChatGPT Dots cost? At least $100 a month. ChatGPT Pro comes in $100, $200 and $500 tiers, and all three include Dots. A Business Premium seat costs $100 a user a month on annual billing or $125 month to month (OpenAI). Some launch coverage got this wrong: BGR reported Dots only on the $200 tier, but The Verge tested it on the $100 one (The Verge, 2 October 2026), and CBS News gave $20 a month for Business Premium, which is the Standard seat's price. OpenAI says it will publish usage terms for each plan after the first month.

What's the difference between ChatGPT Dots and Muse? Both are always-on agents with their own cloud computer and a browser you can watch and take over, and both ask before they buy. Muse is free up to a weekly limit for adults in the US and Canada, works in its own apps and on WhatsApp, and is built for personal errands and small business. Dots costs at least $100 a month, lives inside ChatGPT, connects to Codex, and leans toward work: The Verge's hands-on found it stronger at work tasks than at errands. The Verge also asked whether Dots can compete with free (The Verge, 1 October 2026). For personal errands in North America, Muse is the cheaper answer.

Is ChatGPT Dots safe? OpenAI has published detailed guardrails. Password changes and money transfers always go back to you, a separate Auto-review model checks emails and other sharing before they go out, and background research is read-only. The open questions are memory you can't edit one item at a time, a model-improvement setting on personal plans that covers your dot's conversations and actions, and the bugs users reported in its first week. It also launched a day after OpenAI apologised for one of its agents accessing Australian government systems.

Is ChatGPT Dots any good? Early reviews are positive about work. Casey Newton estimated his dot did about two hours of work for about 15 minutes of his time (Platformer, 30 September 2026). The Verge's hands-on found it stronger at work tasks than at errands. The criticism is about access and cost: ZDNET called the cost structure open-ended and underdefined, Benedict Evans called the launch confused, and users in Europe and the UK have complained that they can't get it on Pro.

Can you text ChatGPT Dots? Only in a limited beta. As of October 2026 texting is open to some Pro users in the US, through a third-party provider, and not to Business or Enterprise workspaces, while OpenAI's marketing pages still say texting is coming. No dots page mentions WhatsApp. Separately, the ChatGPT desktop app can read and send your iMessages from an Apple silicon Mac (how that works); OpenAI hasn't said whether a dot can use it. To text an assistant today, Arlo works in iMessage, SMS and WhatsApp, Pally in iMessage, RCS, SMS and Telegram, and Instinct in iMessage, SMS and WhatsApp with an invite. More: the best AI assistants you can text.

Can ChatGPT Dots make phone calls? You can call your dot by voice inside the ChatGPT app, though users reported those calls failing to connect in its first week. Your dot can't call you at launch, and OpenAI doesn't list phone calls to businesses or people on your behalf. Arlo places calls to people and businesses, works through phone menus and waits on hold. Pally includes calls, and Muse and Instinct call businesses in beta or early access. More: AI assistants that can make phone calls.

Is ChatGPT Dots available in the UK and Europe? On ChatGPT Pro, not yet. At launch Pro excludes the EEA, Switzerland and the UK, and OpenAI hasn't given a date. Business Premium seats include Dots in every supported ChatGPT region. Sam Altman told NBC News the EU process "takes longer, and sometimes unpredictably longer".