Grok Bot vs Muse vs OpenAI Dots: Which Personal Agent Should Run Your Business Personas?

Updated: November 12, 2026 · By Brian Arrington (Founder, AI Consultant and Business Coach)*

The short answer

Grok Bot is best for people who want a team of named Bots, each with its own job, that can be shared as templates.

Muse by Meta is best for people who want one personal agent whose personality lives in an editable Soul.md.

OpenAI Dots is best for ChatGPT Pro or Business Premium users who want one always-on agent with clear action rules.

All three can run a BotPlaybooks persona.

I am not going to name a winner. All three are personal agents, and each one is built around a different idea of how you work. This comparison is written from the persona side: where your agent's lasting role and rules live, how you load a ready-made persona, and how each one asks before it acts. Every fact comes from the maker's own pages, listed in Sources.

Grok Bot vs Muse vs OpenAI Dots at a glance

Access and plans change often, so check each maker's own plans page before you decide: Grok Bot plans, Muse subscriptions and OpenAI Dots plans. Access in this table was read on October 20, 2026; the Grok Bot row was rechecked November 1, 2026.

Where a persona lives in each one

A persona is the lasting part of an agent: who it is, its job and what it should never do. Each tool keeps it somewhere different.

Grok Bot: use the plus button

Grok Bot does not use a SOUL.md file. In your main Grok Bot, use the plus button to give it the Markdown file. Tell it to take that file and build the persona as a new agent. One-off task details go in your later messages. Today's work goes in the chat.

Muse: Soul.md, Identity.md and Memory.md

Muse is the closest of the three to the SOUL.md idea used by Hermes and OpenClaw. Meta's help center describes Soul.md as holding "core truths, boundaries, and personality," and lists Identity.md and Memory.md beside it. You can edit the files in the app or ask Muse to change itself in chat, and "Go back to your default personality" resets it [4][9].

OpenAI Dots: instructions in the conversation

A dot has no persona file to paste into. You give it instructions in the conversation, and you change the request by replying. Its name and avatar are editable in the dot's profile [6]. For actions, OpenAI adds Custom rules under Settings > Personalization > Custom rules, which control what the dot may do on its own [7].

How each one asks before it acts

This is the part I care about most, because every BotPlaybooks persona follows one rule: the agent drafts, and you approve anything sent, posted or paid.

OpenAI Dots documents this best. OpenAI says a dot runs an automatic review before actions that could affect your accounts or share information. Custom rules give you four choices for a specific action: take action without asking, take action when you say so, ask before taking action, or hand off to you. One line on OpenAI's controls page is worth knowing: "Asking your dot to draft replies doesn't give it permission to send them." [7]

Grok Bot is "designed to complete work while keeping sensitive inputs and consequential actions under your control." "Auto Review checks each action your Bot is about to take," and "Auto Review is on by default." When an action needs approval, you get a card. "Allow once: the Bot does the action now. No rule is saved." "Deny: the Bot doesn't do it." There is also Always allow, which saves a matching rule. A card started from your chat doesn't expire on a timer, but "When a routine, an incoming message, or other background work raised the card, it expires after 30 minutes." For logins, "Bots sign in and check out without seeing secret values." You take over only for steps like single sign-on, passkeys, CAPTCHAs and bank card checks. You can also set Ask first rules, and if rules conflict, "Ask first wins." xAI's own advice lines up with ours: "Keep sending, publishing, purchasing, deletion, and production changes behind approval." [13]

Muse checks with you "before sensitive actions like payments or purchases," according to Meta. Connector permissions let you choose how often it asks, and the Always ask setting means "your Muse will ask for permission before any action." When it asks, you can allow once, for the task, for the site, always, or deny. Meta also says you are responsible for "approving its actions." [14]

Whatever the tool does on its own, put the approval rule in the persona too. In Grok Bot that means the persona you load by giving the Markdown file to your main Grok Bot and telling it to build a new agent. In Muse it means Soul.md, plus the Always ask setting if you want Muse to check before every action. In a dot it means your instructions, backed by a Custom rule such as "Ask before taking action" for sending messages. Our recommendation is to keep that rule to one clear line, such as "Draft only; ask me before anything is sent, posted or paid," so it is easy to find and hard to miss.

One agent or a team?

This is the biggest practical difference.

Grok Bot runs many named Bots, and all of them share one persistent cloud computer [1][2]. If you want a Chief of Staff and a Personal Assistant as two separate agents, each with its own job and conversation, Grok Bot is built for that.

Muse is one personal agent [3]. You can change its personality, but there is one Muse. If you want two roles, you either combine them into one Soul.md or switch between them.

OpenAI says each member can create one personal dot [6]. One dot means one main role, with the rest handled through your instructions.

If you need several roles side by side, start with the tool that runs several agents.

Sharing works differently too. Grok Bot can share a Bot as a template [1]. For a dot, OpenAI documents no sharing feature, and "Only a dot's owner can direct it." A dot can post in Slack or Teams channels you add it to, where other members see its messages, but each person has their own dot [15]. Meta's pages do not document sharing a Muse persona.

Loading a BotPlaybooks persona in each one

Each BotPlaybooks persona is one Markdown file with a system prompt (role, inputs, workflow, output format, guardrails), tools and connectors, routines and an example first task. It is not named SOUL.md out of the box, so you place each part where your tool expects it. The steps below use the agent called "Chief", the Chief of Staff persona, as the example. For the full walkthrough, see the setup guide.

In Grok Bot:

1. In your main Grok Bot, use the plus button to give it the Markdown file.
2. Tell it to take that file and build the persona as a new agent.
3. A Chief of Staff or CEO bot helps. If you have one, tell it that it's in charge of the new persona and that the new persona reports to it.
4. Then send the Example First Task.

In Muse:

1. Back up your current Soul.md, Identity.md and Memory.md text.
2. Paste the persona's identity, role, voice and guardrails into Soul.md.
3. Update Identity.md, which Meta describes as "Information about your Muse (examples: its name, creature, vibe, or tag line)." [4] Give it the persona's name, for example Chief, and a short tagline for the role. Keep the job itself in Soul.md.
4. Send the example first task in chat.
5. If you want to undo it, ask Muse to "Go back to your default personality."

In OpenAI Dots:

1. Open your dot in ChatGPT.
2. Send the persona as working instructions in the conversation: role, inputs, workflow, output format and guardrails.
3. In Settings > Personalization > Custom rules, set "Ask before taking action" for sending messages and anything else you want to approve.
4. Send the example first task.

Good roles to try next: the agent called "Daybreak" (Daily Planner persona), the agent called "Errand" (Personal Assistant persona), the agent called "Docket" (Inbox Triage Assistant persona) and the agent called "Book" (Scheduling Coordinator persona). There are setup notes for Muse and Dots on the site.

What about OpenClaw and Hermes?

OpenClaw and Hermes are self-hosted agents, and both use a SOUL.md file for the agent's identity [16][17], so a persona maps to them much the way it maps to Muse.

FAQ

What is the difference between Grok Bot, Muse and OpenAI Dots?

Grok Bot runs a team of named Bots. You load each persona by using the plus button in your main Grok Bot to give it the Markdown file and telling it to build the persona as a new agent. Muse is one personal agent whose personality lives in Soul.md, beside Identity.md and Memory.md. OpenAI Dots gives you one always-on agent, a dot, that follows instructions in the conversation plus optional Custom rules for actions.

Does Grok Bot use a SOUL.md file?

No. Grok Bot does not use a SOUL.md file. To load a persona, use the plus button in your main Grok Bot to give it the persona's Markdown file. Tell it to take that file and build the persona as a new agent. A Chief of Staff or CEO bot helps. If you have one, tell it that it's in charge of the new persona and that the new persona reports to it. Then send the Example First Task. One-off task details go in your later messages.

Does Muse use SOUL.md?

Yes. Muse uses a file called Soul.md, which Meta describes as holding "core truths, boundaries, and personality." It sits beside Identity.md and Memory.md, and you can edit them in the app or ask Muse to change in chat. Asking it to go back to its default personality resets it.

How do I give an OpenAI dot a job?

Give it clear instructions in the conversation: the role, the inputs, the steps and what you want back. Reply in the same conversation to change the request. For actions, set Custom rules in Settings, such as "Ask before taking action." OpenAI notes that asking a dot to draft replies does not give it permission to send them.

Who can use OpenAI Dots?

OpenAI says Dots is available to eligible Pro users aged 18 and older outside the EEA, United Kingdom and Switzerland, and to Business Premium users in supported regions. Enterprise, Edu and Healthcare are generally available in eligible workspaces. Each member can create one personal dot. Access changes often, so check OpenAI's plans page before you decide.

Can I run more than one agent?

In Grok Bot, yes: you can run many named Bots, each with its own job, on one shared cloud computer. Muse is one personal agent. OpenAI says each member can create one personal dot. Pick based on how many roles you need.

Do BotPlaybooks personas work in all three?

Yes. Each persona is one plain Markdown file, so you can place its parts wherever each tool expects them. In Grok Bot, use the plus button in your main Grok Bot to give it the Markdown file and tell it to build the persona as a new agent. In Muse, the role goes in Soul.md and the name in Identity.md. In OpenAI Dots, you send it as instructions and add Custom rules for approvals.

Give your agent a job today

• Browse the persona library and pick a role for your agent. Every persona in the library has a job, inputs, outputs and guardrails.
• Get a free persona and try it in Grok Bot, Muse or a dot this week.
• Need several roles at once? See the packs.

Every persona drafts. You approve anything sent, posted or paid.

More guides coming soon.

Sources

Official maker pages only, checked Oct 16 to 20, 2026. Grok Bot pages [2] and [13] rechecked Nov 1, 2026, and [1] rechecked Nov 2, 2026.

1. Grok Bot, Bots: https://docs.x.ai/grok-bot/bots
2. Grok Bot, overview: https://docs.x.ai/grok-bot/overview
3. Muse, personal AI agent: https://www.meta.com/help/artificial-intelligence/1303670544995562/
4. Muse, Soul.md, Identity.md and Memory.md: https://www.meta.com/help/artificial-intelligence/2225571704857152/
5. Muse, cloud VM and Secure Credentials Store: https://www.meta.com/help/artificial-intelligence/1047255454427887/
6. OpenAI, getting started with your dot: https://help.openai.com/en/articles/20001530-getting-started-with-your-dot
7. OpenAI, Dots controls: https://learn.chatgpt.com/docs/dots/controls
8. Muse, get started: https://www.meta.com/help/artificial-intelligence/1331373868832401/
9. Muse, default personality: https://www.meta.com/help/artificial-intelligence/995796179982326/
10. Grok Bot plans: https://cursor.com/help/grok-bot/plans
11. Muse subscriptions: https://www.meta.com/help/subscriptions/1021145227643680/ and https://www.meta.com/help/subscriptions/1625680452306909/
12. OpenAI Dots docs and plans: https://learn.chatgpt.com/docs/dots
13. Grok Bot, approvals, security and privacy: https://docs.x.ai/grok-bot/approvals-security-and-privacy
14. Muse, privacy and security: https://www.meta.com/help/artificial-intelligence/1385290430137537/
15. OpenAI, manage dots in ChatGPT workspaces: https://help.openai.com/en/articles/20001554-manage-dots-in-chatgpt-workspaces
16. OpenClaw, SOUL.md template: https://docs.openclaw.ai/reference/templates/SOUL
17. Hermes Agent, personality and SOUL.md: https://hermes-agent.nousresearch.com/docs/user-guide/features/personality

AI can make mistakes. Always review outputs before acting. Personas are templates, not legal, financial, tax, medical or HR advice.

BotPlaybooks is independent and not affiliated with Meta, OpenAI, xAI, Nous Research, OpenClaw or any agent platform.

Brian Arrington is an AI consultant and business coach focused on agentic AI. He founded BotPlaybooks and Arrington Consulting LLC. He is an Air Force veteran and the 2025 Georgia Vetrepreneur of the Year.