ChatGPT Dots vs Grok Bot vs Meta Muse

Snapshot

ChatGPT Dots (OpenAI Dots), Grok Bot (xAI / Cursor), and Meta Muse are all always-on personal agents with a cloud computer. They differ most in how you enter them, who they are built for, how they schedule work, and how privacy and approvals are framed. This page sticks to publicly documented differences so you can match the product to the job.

Use this guide with our getting started, features, integrations, and pricing pages when you are evaluating Dots specifically.

What each product is

| Product | Official framing | Model / runtime called out | | --- | --- | --- | | ChatGPT Dots (OpenAI Dots) | Always-on agent in ChatGPT that keeps work moving between conversations and asks when a decision needs you | Powered by GPT-6 Astra; cloud computer + browser; can start or manage ChatGPT Work / Codex tasks | | Grok Bot | Named AI teammates you message like colleagues; finish multi-step work in real tools | Persistent cloud computer (browser, filesystem, terminal); connectors/MCP plus computer use; multi-Bot coordination | | Meta Muse | Personal AI agent for everyday and small-business goals: messages like a person, acts across apps | Powered by Muse Spark; runs on Muse Secure VM with a dedicated browser; Sentinel approval layer |

Dots sit inside the ChatGPT product surface and can hand coding or broader computer tasks into Work or Codex. Grok Bot is a dedicated Bot roster with parallel teammates on one shared account computer. Muse is positioned as a consumer / small-business personal agent with WhatsApp and Meta-app reach.

Platforms and how you start

| Axis | ChatGPT Dots | Grok Bot | Meta Muse | | --- | --- | --- | --- | | Primary entry | ChatGPT desktop app or desktop browser (chatgpt.com/dots); same Dot later in the mobile app when supported | Dedicated Grok Bot desktop app (macOS, Windows, Linux) + iOS/Android companion | Muse app (iOS, Android), muse.ai; WhatsApp messaging; AI glasses “coming soon” | | Mobile web create | Not supported for setup | N/A (native apps) | Web UI available (muse.ai) | | Messaging channels | ChatGPT chat/call; Slack and Microsoft Teams; phone texting labeled coming soon | Message Bots in the Grok Bot apps (type, dictate, voice chat) | Muse app and WhatsApp (Meta announcement) | | Typical first step | Create one Dot, connect apps, assign a responsibility | Create a named Bot, grant access as asked, hand off a multi-tool task | Tell Muse a goal or task; connect apps; approve sensitive actions |

For Dots setup details, see getting started and access.

Scheduling and routines

• Dots: Work continues between conversations without requiring a fixed schedule for every follow-up. You can still request timed check-ins (include time zone and an end date) and manage recurring items under Scheduled. Proactive research can run in the background; follow-up actions still need permissions.
• Grok Bot: Skills describe how to do work; routines assign a workflow to a Bot on a schedule or supported event. You can teach a workflow once (including demonstration where available) and reuse it.
• Muse: Keeps working after you close the app and returns when something changes or approval is needed. Personal goals become plans it advances over time; small-business flows emphasize ongoing growth and ops goals.

Apps and computer access

All three give the agent a cloud computer with a browser so work can continue while your laptop is closed.

| Capability | ChatGPT Dots | Grok Bot | Meta Muse | | --- | --- | --- | --- | | Cloud computer | Per-Dot cloud computer and browser | One shared persistent cloud computer for all of your Bots | Dedicated Muse Secure VM per person | | Local computer | Optional connect of one personal computer (must be online with ChatGPT app open); separate from Codex / Work Sync | Optional local execution with Ask every time / Always / Never; team admins can cap it | Agent work is VM-centric; clients connect to your VM | | App connections | ChatGPT plugins (e.g. Gmail, Drive, GitHub) with shared plugin permissions | Connectors / MCP where available; computer use for sites without a clean API | Built-in connectors (Meta apps + third parties listed for small business) and custom connectors | | Multi-agent | Primary Dot can run background agents and delegate Work/Codex tasks | Multiple named Bots in parallel; group chats and handoffs | Subagents / swarms described in Meta’s safety write-up |

Dots keep cloud sessions separate from your personal browser; website sign-in uses a private request or Take over. Grok Bot docs warn that separate Bots are not a security boundary because they share logins on one computer. Muse stores credentials outside the agent runtime and uses Sentinel for egress and connector actions.

Practical Dots app setup: integrations.

Privacy, safety, and approvals

| Theme | ChatGPT Dots | Grok Bot | Meta Muse | | --- | --- | --- | --- | | Default protections | Built-in safeguards + Auto-review vs instructions, Custom Rules, and safety requirements | Auto Review rules (Ask first / Allow automatically); human takeover for passwords, 2FA, CAPTCHAs, payments | Separate Sentinel agent as sole permission authority for connectors and network egress; credential surrogation so Muse does not see real tokens | | User controls | Custom Rules; Pause / Activity / Scheduled stop controls; plugin and computer permissions | Bot description boundaries; Auto-review table; local-computer policy; connector least privilege | Per-connector read/write and fine-grained actions; audit trail; “forget” memory; training opt-out | | Data / training notes | Lives in ChatGPT account / workspace controls (see OpenAI Dots privacy FAQs) | Uses Cursor authentication and privacy settings; Legacy Privacy Mode not supported | Conversations/VM data not shared with Meta ad systems; sanitized trajectories may train unless you opt out; Confidential VM planned later in 2026 | | Isolation note | Cloud computer separate from your devices | Account-shared Bot computer | Per-user dedicated Secure VM; runtime cell isolated from credential store |

None of these products claim agents never make mistakes. Official docs for all three stress reviewing consequential actions (send, purchase, publish, production changes).

Pricing and availability (as published)

| Product | Access pattern (public docs) | Pricing notes | | --- | --- | --- | | ChatGPT Dots | Rolling out to eligible Pro 100/200/500 (18+, outside EEA/UK/Switzerland), Business Premium worldwide, Enterprise (admin enables Use dots Beta; off by default) | First Dot included with eligible plans; no separate Dots SKU; Dot chats do not count toward ChatGPT limits; Work/Codex tasks use their own limits; launch-month deeper-work promotion | | Grok Bot | Early beta via eligible Cursor paid/Teams plans and/or linked SuperGrok subscriptions; desktop + mobile apps; Enterprise rolling out | Included with qualifying plans; own usage pool (often weekly); on-demand usage may apply—check current Cursor / xAI access pages | | Meta Muse | US (and Canada in later Meta posts) on iOS, Android, muse.ai; WhatsApp; glasses later | Free for most everyday use, with paid subscriptions for higher usage; Muse for Small Business expands connectors |

Eligibility and prices change. Confirm on the vendor’s current access page and, for Dots, our pricing and availability guides.

Best-fit scenarios

Choose ChatGPT Dots (OpenAI Dots) when you already live in ChatGPT, need one primary always-on coordinator that can reach plugins plus Work/Codex, and want workspace admin controls for Business/Enterprise.

Choose Grok Bot when you want a team of named Bots that collaborate, share one computer for handoffs, learn skills/routines from demonstration, and you already have Cursor or SuperGrok access.

Choose Meta Muse when you want a consumer-friendly personal agent with WhatsApp entry, strong Meta/small-business connector coverage, and Meta’s Secure VM / Sentinel privacy story—especially if you are in the published US/Canada rollout.

For how Dots relate to Codex and ChatGPT Work inside OpenAI’s stack, see Dots vs Codex.

Quick comparison checklist

1. Surface: ChatGPT vs dedicated Grok Bot apps vs Muse/WhatsApp.
2. One agent vs many: Primary Dot vs multi-Bot roster vs personal Muse (+ subagents).
3. Computer model: Per-Dot cloud PC vs one shared Bot PC vs dedicated Muse Secure VM.
4. Scheduling: Conversational continuity + Scheduled vs skills/routines vs goal-driven background work.
5. Admin / region: ChatGPT plan+region+Enterprise toggle vs Cursor/SuperGrok entitlement vs Muse geo rollout.
6. Privacy posture: Custom Rules + Auto-review vs Cursor privacy + Auto Review vs Sentinel + credential isolation (+ upcoming Confidential VM).

Sources

Claims above are drawn from OpenAI Meet dots / Dots feature pages and Help Center workspace docs; xAI Grok Bot overview, launch post, FAQ, skills/routines, and approvals docs; and Meta’s Muse introduction, safety research post, and Muse for Small Business announcement. Re-check those pages before purchase or rollout decisions.