Claude Duet vs dots vs Muse vs Grok Bot (October 2026)

As of October 2026, Claude Duet, OpenAI dots, Meta Muse and Grok Bot are all personal agents that work on a cloud computer of their own, and none of them shares memory, files or tasks with the others. Whether you are weighing Claude Duet vs dots or all four, the practical differences are which subscription you need, how much control you get over memory, and whether more than one person can use the agent. There is no single best one: the right pick is usually the agent attached to the account you already pay for.

How do Claude Duet, dots, Muse and Grok Bot compare?

Each lives inside its maker's account and connects to apps its own way. Cells come from the maker's pages unless marked otherwise.

What is the difference between Claude Duet and OpenAI dots?

Duet extends your Claude account; a dot extends your ChatGPT account. Each inherits memory and apps from its parent, and neither sees the other.

Duet checks your connected accounts "every few hours", and those background checks count toward your Claude limits (Anthropic). A dot reaches you in ChatGPT, Slack or Teams; chats with it do not count toward ChatGPT limits, but tasks it runs in Codex or ChatGPT Work do (OpenAI).

They also split memory differently. Duet keeps one main thread, and the separate channels you start share the same memory (Claude help). Each task you give a dot has its own conversation and does not automatically receive everything you have discussed with your dot (OpenAI help).

Muse vs dots: which gives you more control over memory?

Muse shows you its memory as files; a dot does not let you see individual memories.

Meta says you can "inspect, edit and download" Muse's memory files (Meta). OpenAI says you cannot view or delete individual dot memories; to clear them you reset or delete the dot (FAQ). With both, disconnecting an app does not erase what the agent already picked up (for Muse, per Meta's help center as quoted by PortEden).

If memory you can open is the deciding factor, an open agent framework is another option: OpenClaw "remembers things by writing plain Markdown files in your agent's workspace" (docs). See how OpenClaw memory works.

Grok Bot vs Muse: which works for a team or a household?

Grok Bot is the only one of the four with a documented team feature. Muse documents no way for two people's Muses to work together.

A Team Bot is "one Bot that your whole team talks to" (docs), with a chat per teammate and team memory they all read. Sharing a personal Bot only creates a template copy without "your computer, logins, or conversation history" (docs). As one user on Cursor's forum put it: "Copies do not share memory, files, logins, or conversation history" (forum).

Which personal AI agent is free, and what does each cost?

Only Meta Muse is free to start: Meta says it is "free for most of what people need, with subscription plans for people who want to do more" (Meta). Duet, dots and Grok Bot each come with a paid plan you may already have.

• Claude Duet: included with Pro and Max at no extra cost during the preview. Pro is $20 a month billed monthly (pricing). Duet uses your plan's limits, and when you are admitted you get one full limit reset, available until November 30, 2026 (Claude help).
• OpenAI dots: your first dot is included in Pro or Business Premium (OpenAI). Pro starts at $100 a month (pricing).
• Grok Bot: included with every paid individual Cursor plan and the Cursor Teams plan (Grok Bot FAQ), or with X Premium ($30 a month), Super ($100) and Ultra ($200) (x.ai). Usage resets weekly.
• Meta Muse: the paid plan names and prices (Power and Maximum) are press-reported by TechCrunch, not on Meta's pages.

Which personal AI agent is best for you?

None of them is best for everyone. Each suits a different starting point:

• Claude Duet if you are in the US on Claude Pro or Max, your apps are connected in Claude, and you want a morning brief.
• OpenAI dots if you are on ChatGPT Pro (outside the EEA, Switzerland and the UK) or Business Premium and work in Slack or Teams.
• Meta Muse if you are in the US, want to read and edit what the agent remembers, and are happy to build connectors through chat.
• Grok Bot if you want several named Bots with separate roles, or one Bot a team shares.

A frequent question since the Duet launch is whether it is worth switching from Grok Bot or a dot. For most people it comes down to which subscription and connected apps they already have.

What about Gemini Spark?

Gemini Spark is Google's entry: "a 24/7 personal AI agent designed to proactively manage tasks" (Google, May 19, 2026). Google's help center says Spark tasks need a Google AI Pro or Ultra subscription and a personal Google Account (not work or school), you must be 18 or over, and tasks are not offered in the EEA, Nigeria, Switzerland or the UK. Spark can use Personal Intelligence, including memory of your past Gemini chats, and is designed to ask before actions such as sending messages or making purchases.

It is not in the table because Google's pages do not describe its computer, custom MCP connections or sharing in the same terms. Tokenrip has no route for Gemini today: Gemini custom apps need dynamic client registration, which Tokenrip's OAuth server does not offer.

Can these agents share memory or work with each other?

No. As of October 2026 each agent's memory stays with its maker: a dot receives memories from ChatGPT and makes its own, Duet draws on your Claude account, and Muse keeps its memory on its own computer. Grok Bots can message each other directly and in group chats, but that stays inside Grok Bot. Grok chat on grok.com is a separate product with its own memory (see Grok memory).

Users ask for it; one OpenAI community thread is titled "Multiple specialized dots with shared context and clear task ownership" (OpenAI Community). If you run two of these agents, the shared context has to live outside all of them, in something each one connects to (see MCP vs A2A vs Tokenrip).

How do you share work between personal agents without Tokenrip?

Keep one shared document that every agent reaches through its own connectors, with your decisions and rules at the top. Duet, for example, lists Google Drive among its connectors; check each agent's list before you pick a home.

This is enough when one agent does most of the writing. It gets harder when two agents edit the same file, when you need to know which agent changed what, or when a job must not be done twice. Without a shared place, you become the go-between, re-explaining the same context to each agent.

How do you connect each personal agent to Tokenrip?

Tokenrip is the shared workspace for your AI agents. Claude, ChatGPT, Codex and Grok bring their work here. You approve what needs you and see the rest get done.

1. Connect your first agent. Paste into any agent you use: Connect me to Tokenrip. Fetch https://tokenrip.com/setup and follow it. You give an email address and a six-digit code. Duet, Muse and Grok Bot each have a cloud computer, so you can also try this line in them directly; it uses the Tokenrip CLI and is unconfirmed on those three.
2. Claude Duet. Add Tokenrip in Claude as a custom connector with the URL https://api.tokenrip.com/mcp, then ask Duet to open your Tokenrip workspace. Duet works with the connectors you connected in Claude, but Anthropic has not documented custom connectors in Duet yet; if Duet cannot see Tokenrip, use it from a Claude chat or cloud Cowork, which share the same Claude memory (steps).
3. OpenAI dots. Add Tokenrip to ChatGPT on the web with OAuth (how), then ask your dot to use it. OpenAI's help center adds custom MCP servers under Plugins > + > Add custom MCP server and gives Pro read/fetch permissions only. OpenAI does not document whether a custom MCP plugin reaches a dot, so this is unconfirmed.
4. Meta Muse. Ask Muse to build a custom connector to the same URL and enter a Tokenrip key only in its secure credentials prompt, never in chat (see how to build a Muse custom connector). Meta says Muse can write its own connectors, but this is unconfirmed with Tokenrip.
5. Grok Bot. Tell your Bot "Add this MCP server: https://api.tokenrip.com/mcp" and sign in on the Tokenrip page. Do not paste a key or ask a personal Bot to send a key header: Cursor staff wrote on 2026-09-17 that "there's no safe way in Grok Bot to set a custom HTTP header" and that connectors authenticate only via OAuth (thread). Tokenrip supports OAuth through Client ID Metadata Documents, but compatibility with Grok Bot is not yet confirmed. On a Team Bot, the owner can store a Tokenrip key as the Bot's encrypted secret (steps).
6. Pin decisions and invite people. Every agent reads pinned documents first when it loads the workspace. People you invite join free and bring their own agents.

Example. Duet's morning brief spots a supplier email that needs a reply and saves a note to the workspace. Your Grok Bot starts the "reply to supplier" task (which it holds, so no other agent takes it) and files the draft as a proposal; you approve it, edit it or skip it. That evening your partner's Muse loads the workspace and sees what changed since it last looked.

Tokenrip does not read any agent's native memory; agents read and write the workspace when connected and asked to.

Try it in one step. Paste this line into the agent you want to connect. It reads the setup page and does the rest; you give it your email address and a six-digit code.

Connect me to Tokenrip. Fetch https://tokenrip.com/setup?ref=personal-ai-agents-compared and follow it.

Related reading

• Claude Duet connectors
• OpenAI dots memory
• Meta Muse custom connector
• Grok Bot memory

Connect your personal agent to Tokenrip