OpenAI Dots or Grok Bot? Comparing Always-On AI Agents

Autumn 2026 may be remembered as the season the "chatbot" era ended. In August, xAI (now SpaceXAI) launched Grok Bot; at the end of September, OpenAI announced dots. Both promise the same thing: not an assistant that answers you, but a teammate with its own computer that finishes the work for you, even while you sleep.

On paper they look almost like twins. Look closer, though, and their philosophies diverge. In this post we put the two side by side using their official announcements and documentation.

In short:

• Both are always-on agents with their own cloud computer that you hand work to by message.
• Dots start with one personal agent and build safety around detailed "Custom rules".
• Grok Bot is designed as a team from day one: several Bots work in parallel and hand work to each other.
• Grok Bot's desktop app supports macOS, Windows and Linux; dots are rolling out to Pro users outside the EEA, Switzerland and the UK.

What do both products promise?

The two companies tell a strikingly similar story. OpenAI calls dots "remarkably capable, always-on agents"; SpaceXAI introduces Grok Bot as "your team of always-on agents" that "keep working 24/7".

The shared idea: a classic chatbot runs when you type and stops when you close the window. These new agents take on a goal, plan the steps themselves, work in real apps and come back only when a decision needs you. A line from the Grok Bot announcement captures it well: most AI gets you to 90%, and the hard part is the last 10%, where the work lands in the actual tool a human would use.

Similarities: two takes on the same idea

They have their own computers

In both products the agent owns a computer in the cloud. Each dot has its own cloud computer and browser and can write and test code when needed. Grok Bots work on a persistent cloud computer with a browser, filesystem and terminal. The result is the same: close your laptop and the work continues.

You send a message, not build a workflow

Both companies push back on "setting up automations". Grok Bot's docs describe setup as "a message, not a workflow builder". With dots, you write to the agent as you would to a colleague. No drag-and-drop flowcharts.

They learn over time

Dots learn from your feedback and pick up your preferences and way of thinking. Grok Bots remember conversations, learn how you like things done and "get sharper the more you work together".

They come back to a human at critical steps

Neither promises full autonomy. Dots turn to you for decisions that need your judgment; Grok Bots finish jobs end to end and "only come back when something needs your approval".

Differences: where the philosophies split

One assistant or a team?

This is the core difference. Dots start with one personal agent. OpenAI says you can start with a single dot today and expects teams of dots to work together over time; adding more dots comes "in the future".

Grok Bot is a team from day one. SpaceXAI says its staff often run several Bots in parallel, with a chief of staff Bot on top and a specialist for each lane: inbox, expenses, recruiting, bug fixes. Bots can message each other directly, coordinate in group chats and pass ownership of a task. In one example from the announcement, an engineering Bot reproduces a bug in the UI, files the ticket and hands the fix to a debugging Bot.

Computer sharing

In dots, each dot runs on its own cloud computer. In Grok Bot, all the Bots on your account share one cloud computer: files, browser sessions and app logins are shared. That makes handoffs easy, but the docs warn that anything you place on that computer should be treated as available to every Bot you run. Isolation between users is strict. Each Bot gets its own screen on the shared computer and runs one computer-use task at a time.

Teaching style: tell or show?

One of Grok Bot's standout features is learning by demonstration. You ask a Bot to follow along while you do a job; it watches the steps, saves them as a routine or skill, takes your corrections and runs it on its own next time, on a schedule if you like. With dots, learning happens mostly through feedback and memories formed from connected apps.

Safety and control model

OpenAI documents dots' safety side in detail. With Custom rules you choose, per action, between "take action without asking", "take action if pre-approved", "ask before taking action" or "hand off to you". Actions that could affect your accounts go through automatic review. Proactive research while you're away uses only read-only tools. Dots can use saved passwords on supported sites without exposing them to the model, and a monitoring system can pause a dot if it detects a safety issue.

Grok Bot also has approvals, Auto Review, network policy and logging. When a Bot hits a site that blocks automation or requires a human step, it hands that step to you rather than working around it. For enterprises there are admin controls, identity management and private network options.

Platforms

Pricing and access

Dots are rolling out to ChatGPT Pro and Business Premium users in eligible markets; Enterprise users can try the beta once an admin enables it. The first dot is included at no extra cost, and conversations with it don't count toward ChatGPT usage limits. Pro access excludes the EEA, Switzerland and the UK.

Grok Bot launched in beta for SuperGrok, SuperGrok Plus and SuperGrok Heavy subscribers. An interesting detail: it's also included with every paid individual Cursor plan and the Cursor Teams plan, and the computers Bots work on run in Cursor's cloud. Grok Bot has its own usage, separate from your Grok and Cursor plans, which resets weekly. Enterprise customers can join a waitlist.

Which one fits whom?

There's no clear winner; the two serve different working styles.

Dots may suit you better if:

• You already use ChatGPT Pro or Business Premium and your team lives in Slack or Teams,
• You want one strong personal assistant,
• Detailed, per-action safety rules are a priority,
• You lean on the Microsoft ecosystem (OpenAI is working with Microsoft to bring specialist dots to Microsoft Agent 365).

Grok Bot may suit you better if:

• You want a small team of agents running several jobs in parallel,
• You'd rather show a job once than describe it,
• You use a Linux desktop or already pay for Cursor,
• You work a lot with tools that have no API or connector.

The shared risk: how much should you trust agents?

Both products share a risk: the agent works in your real accounts with your real data. A web page or email the agent reads may contain hidden instructions. OpenAI says dots' built-in safeguards help protect against malicious instructions, but both companies acknowledge mistakes can happen.

The practical advice is the same for both: give agents only the access they need at first, put spending and outbound sharing behind approval, and review the output carefully in the first weeks. If you're setting up dots, our OpenAI dots guide walks through these rules step by step.

Frequently asked questions

Do dots and Grok Bot do the same job?

The core idea is the same: always-on agents with their own cloud computer that take work by message. The difference is the approach. Dots start with one personal agent and a detailed rules system; Grok Bot is built around teams of Bots that talk to each other and work in parallel.

Which one is safer?

Both offer approvals, automatic review and monitoring. OpenAI documents dots' safety model (Custom rules, read-only proactive research, passwords hidden from the model) in more detail. With Grok Bot, keep in mind that all Bots share one computer and everything on it is available to all your Bots.

Does it make sense to use both?

It's possible, but unnecessary for most people and small businesses. You'll usually get more by picking the ecosystem you already spend time in (ChatGPT/Microsoft or Grok/Cursor) and going deep.

Where does Meta's agent fit?

Meta's Muse agent is part of the same race; we looked at its features in our Meta AI agent Muse post. The press widely read OpenAI's dots launch as a direct answer to Meta.

Always-on agents are changing how businesses relate to software: you no longer ask AI questions, you hand it work. If you want agent integrations built around your own processes, reach us through our enterprise software development page.

Sources