Grok Bot vs Dots: Where the Feature Sets Actually Diverge

Strip away the branding and Grok Bot and Dots start from the same sentence: give an AI agent its own computer, let it keep working after the human logs off, and connect it to the tools people already use. It's a genuinely convergent idea — xAI, OpenAI, and Meta all arrived at some version of it within weeks of each other. But "same premise" is where the similarity stops. Once you look at how each product actually gets things done, Grok Bot and OpenAI's newly launched Dots read like answers to two different questions.

The credential question is the real divide

Grok Bot's pitch, laid out on x.ai's own announcement, is refreshingly blunt: "Bots have their own computer. They sign into the tools you already use and work across apps, inboxes, and more." That's not a metaphor. Grok Bot logs into your CRM, your inbox, and your other work tools using your own credentials — deliberately including tools that have no API or MCP integration at all. It's a "no integration required" pitch: if you can log into it, a bot can too.

Dots, powered by GPT-6 Astra, takes the opposite approach. Each dot gets its own cloud computer and browser, kept separate from the user's machine and accounts unless the user explicitly connects it. Rather than borrowing your login, a dot works through OpenAI's plugin ecosystem — more than 4,000 applications, per OpenAI — and through Custom Rules and an Auto-Review layer that checks consequential actions before they execute. It's an integration-first model: broader reach through official connections, narrower reach into anything that isn't connected.

Neither approach is obviously "safer" in the abstract. Grok Bot's model is more universally capable on day one — it doesn't need a vendor partnership to work with your invoicing tool, it just needs your password. Dots' model is more auditable and constrained — it can only do what the plugin ecosystem and its permission layer allow, which is a smaller universe but a better-fenced one. We go deeper on the security implications of that split in Credential Sharing: The Quiet Risk Behind "Signs In as You" and in the OpenAI-side companion piece, Dots vs Grok Bot vs Muse: Reading the Permission Models Side by Side.

Where each one lives

The two products also differ in where and how you reach them. Dots surfaces through ChatGPT, Codex, Slack, and Microsoft Teams, with phone/text and an audio model described as "coming soon." That's a workplace-messaging-first distribution strategy, reinforced by OpenAI's Agent 365 partnership with Microsoft for enterprise identity and credential provisioning.

Grok Bot is bundled directly into subscriptions people already pay for: SuperGrok, SuperGrok Plus, and SuperGrok Heavy on the consumer side; Cursor Pro, Pro+, and Ultra, plus Cursor Teams Standard and Premium, on the developer side. It runs on desktop and iOS, with enterprise access on a waitlist, and — per xAI — uses "its own usage, separate from your Grok and Cursor plans." That's a distribution strategy built around two existing subscriber bases: people already paying for Grok, and developers already paying for Cursor. It's a smaller starting footprint than Dots' workplace-messaging reach, but a more captive one — you're already inside the product Grok Bot is upselling.

Learning vs. teaming

Both products claim their agents get better with use, but the framing differs. xAI describes bots that "observe actions, save them as routines, and refine execution from corrections" — a bot watches you do something once, and turns it into a repeatable procedure. Multiple bots can run in parallel, communicate independently, and coordinate in group chats, which is the backbone of xAI's department-by-department pitch to businesses (sales, CRM, marketing, HR, finance, engineering — covered across the rest of this series).

Dots leans on a different metaphor: "teams of Dots working together on your behalf," each with a name and a personality the user sets, plus a Proactive Research feature that lets a dot investigate read-only tools in the background even when nobody has tasked it. OpenAI is explicit that this backgrounding is limited to read-only access — a constraint Grok Bot's own materials don't mirror, because Grok Bot's model of "sign in and act" doesn't naturally distinguish read from write the way a plugin permission system does.

Grok Bot borrows your keys and gets things done anywhere you can log in. Dots keeps its own house and gets things done anywhere OpenAI has built a door. Neither is halfway — and the choice you make between them is really a choice about what you're willing to trust an agent with.

The learning curve each product asks users to climb

There's also a difference in what each product asks a first-time user to understand before they can use it well. Grok Bot's "sign in and act" model has almost no conceptual overhead — if you can explain to a new hire how to log into your CRM, you can explain it to a bot in the same sentence. The tradeoff is that the user has to independently reason about scope: what should this bot be allowed to touch, given that it now has the same access you do? Dots inverts that tradeoff. Understanding what a dot can and can't do requires understanding OpenAI's plugin ecosystem, Custom Rules, and the read-only boundary on Proactive Research — more concepts to learn up front, but each one narrows the blast radius by design rather than leaving it to the user to define. Neither learning curve is inherently easier; they're just front-loaded differently, and a business's existing technical sophistication probably matters more than either product's own documentation in determining which one feels more approachable on day one.

What's still unconfirmed on both sides

Some claims deserve a flag before you build a workflow around them. xAI's own Grok Bot page describes its cloud computer only in terms of a browser and file access; a small number of secondary sources describe a fuller "browser, filesystem, and terminal" environment, but that specific combination is single-sourced in the material we reviewed and should be treated as unconfirmed rather than official. Similarly, Dots' pricing beyond the first free dot for Pro and Business Premium subscribers is not disclosed anywhere in OpenAI's launch materials — an open question we take up in Grok Bot and the Cost of Always-On.

The honest takeaway is that both products are still young enough — Grok Bot at under eight weeks, Dots at one day — that "feature comparison" mostly means comparing stated intentions, not battle-tested behavior. What's clear already is the philosophical split: one company built an agent that borrows your identity to reach everywhere at once; the other built an agent that keeps its own identity and reaches only where it's been formally let in. That's the decision every business piloting either product is actually making, whether or not the marketing frames it that way.

Part 22 of 30 in the Grok Bot series. Previously: The dot.com Stunt: xAI's Marketing Play Against OpenAI's Launch. Next: Grok Bot vs Muse: Enterprise Ambition vs Consumer Trust. Browse more builder insights or explore AI skills at aiskill.market.