Grok Bot vs Meta Muse vs OpenAI Dots: Which Personal AI Agent?

In a few weeks of late 2026, the three biggest AI labs each shipped the same idea: a personal agent that doesn't just answer questions but goes and does the work — on its own computer, connected to your apps, running in the background. xAI has Grok Bot, Meta launched Muse (September 8), and OpenAI unveiled Dots at DevDay (September 29).

On the surface they rhyme. Underneath, they make different bets about the thing that actually matters once an agent can touch your email, your calendar, and your card: how isolated is it, and how does it ask permission? This is a hand-reviewed comparison of what each one is, where they genuinely differ, and which fits which person. Figures labeled "reported" are from secondary coverage and worth confirming with the vendor; everything else is from the companies' own docs.

The one-paragraph version

All three run an agent on a cloud computer, connect to your apps, work while you're away, and stop to ask before risky actions. Muse is the consumer pick: a real free tier, the strongest built-in privacy model, and a shopping layer — but US-only at launch. Dots is for people already living in ChatGPT who want always-on agents across thousands of apps and don't mind paying. Grok Bot is the one built around scheduled routines and triggers — the best fit if your need is "do this job every morning," with one important catch about its shared computer.

What each one is

• Grok Bot (xAI) — an agent built around routines: save a reliable task as a skill, then have a Bot run it on a schedule or from an event (a Slack message, a GitHub notification). Powered by grok-4.7. It lives in the Grok/X ecosystem.
• Meta Muse — a consumer personal agent you talk to in the Muse app or inside WhatsApp. You give it a goal; it builds a plan, opens a browser, fills forms, and advances the work, coming back when it needs approval. Powered by Muse Spark (Meta has since added newer Muse Wonder frontier models).
• OpenAI Dots — "always-on" agents that pursue a goal continuously in the background. Each dot runs on its own cloud computer and reaches ~4,000 apps through ChatGPT. Powered by GPT-6 Astra.

The difference that matters most: isolation

This is where they split hardest, and it's easy to miss.

• Muse gives each person's agent its own Muse Secure VM — a dedicated cloud computer with its own browser, where the agent and your connected-app credentials live, walled off so no other agent can reach it.
• Dots similarly gives each dot its own cloud computer and browser.
• Grok Bot does it differently: all of your Bots share one cloud computer tied to your account. xAI's own docs are blunt that you should not use separate Bots as a security boundary — files, browser sessions, and logins on that machine are reachable across your whole Bot roster.

None of these is "wrong," but it changes how you work. With Muse or Dots, spinning up a second agent gives you a clean, separate sandbox. With Grok Bot, a login one routine uses is reachable by every other Bot you own, so the discipline has to come from least-privilege and scoped accounts, not from which Bot owns the task.

How each one asks permission

Every one of these gates risky actions — the question is how firmly.

• Muse runs a separate Sentinel agent on the same machine, kept apart from Muse at the system level: nothing Muse does reaches the internet unless Sentinel approves it. Muse also never sees your passwords or payment details (they go in secure storage it can use but not read), checks with you before sending an email or buying, and shows a complete audit trail. It's the strictest model of the three.
• Dots uses a permission model plus an Auto-review safety layer, and reports back for you to review. Early users praised the results but flagged occasional unapproved actions — a reminder to keep write-access tight.
• Grok Bot has Auto Review rules (Ask first always stops; Allow automatically proceeds unless review finds a reason not to; Ask-first wins when both match). One sharp edge for unattended work: an approval for something that started without you expires after about 10 minutes, so a routine that needs you overnight stalls rather than waits. (We wrote a whole free course on running Grok Bot routines safely around exactly this.)

Scheduling and background work

• Grok Bot is the only one built around explicit scheduling: routines on a clock (as often as every 5 minutes, up to 50 per Bot) or on an event. If your need is "check X every hour" or "when a deploy fails, do Y," this is the natural fit.
• Muse and Dots work in the background too, but as goal pursuit — you hand over an objective and they keep at it, returning when something changes or needs approval, rather than firing on a cron you define.

Where you use it, and where you can't

• Grok Bot — in the Grok/X ecosystem, with Slack and GitHub event triggers.
• Muse — Muse app, WhatsApp, and muse.ai, US only at launch, on iOS and Android, with Meta AI glasses support coming. It pays using Link by Stripe (a one-time-use card that hides your real details, with purchase protections) — a genuine differentiator for shopping tasks.
• Dots — reachable from the ChatGPT app, web, Slack, Microsoft Teams, or a phone call. In beta for ChatGPT Pro, Business Premium, and Enterprise, and not available in the EEA, Switzerland, or the UK at launch.

What they cost

Prices move fast here; confirm before you commit.

• Muse — free for most use, with Power ($20/mo) and Maximum ($100/mo) tiers for heavier usage. The only one of the three with a real free tier.
• Dots — no free tier; one dot is included in the $100/mo ChatGPT Pro plan, and with Business Premium (reported). Enterprise in beta.
• Grok Bot — bundled into paid Grok / SuperGrok subscription tiers (reported from roughly $30/mo at the base, rising with heavier tiers). Verify current tiers with xAI.

Side by side

Which should you pick?

• Choose Meta Muse if you want a consumer personal agent with a free tier, the strongest built-in privacy and approval model, and help with real-world shopping — and you're in the US.
• Choose OpenAI Dots if you already live in ChatGPT, want always-on agents spanning thousands of apps, and the $100/mo Pro plan is fine (and you're not in the EEA/UK/Switzerland yet).
• Choose Grok Bot if your real need is scheduled automation — recurring routines and event-triggered jobs — and you're in the Grok/X world. Just design around the shared-computer model: least privilege, scoped logins, and a plan for overnight approvals.

The honest summary: the model each one runs barely matters for most people. What matters is isolation, how it asks permission, whether there's a free tier, and whether it's even available where you are. Decide on those four, and the choice gets easy.

Dig into each one: Grok Bot, Meta Muse, and OpenAI Dots. New here on agents? Take the free course: Running Personal AI Agents Safely & Cheaply.