Grok Bot review, pricing and verdict

The vendor describes several working patterns. Bots can be run in parallel, with one Bot coordinating the others, and can message each other and share context in threads or group chats. A Bot can also watch a person complete a workflow once, save it as a routine, and run it on its own next time. Memory carries across conversations, so a Bot picks up how its owner likes work done. xAI's own examples are a sales Bot that researches accounts overnight and leaves drafts to approve, an ops Bot that processes invoices arriving in Gmail, and an engineering Bot that reproduces a bug in a product UI and files the ticket. On 2 November it added Data Bot, a Team Bot that its data science team teaches, which answers warehouse questions in the app or in Slack and shows managers each memory it proposes before the team shares it. Those are vendor-reported internal uses, not independent results.

Since our 22 October check, the changelog (version 0.78.1, released 27 October) added a Payments tab in the Vault where a person picks the payment method Bots use first and still approves every purchase, shopping cards with a Buy now button, order tracking, and templates that cancel unused subscriptions, plus Import Memory from another assistant. The model underneath was Grok 4.6 at launch. xAI says Grok 4.7, released 21 September, was trained to understand the Bot app natively, but its post does not say Bot now runs on it.

It suits anyone who needs work done while they are away, not a copilot at their elbow: an operations lead who needs a support queue worked overnight, a founder who wants CRM follow-ups and inbox triage handled before the day starts, or a team already paying for Cursor. Compared with Manus, a hosted research agent, Grok Bot leans on logging into your real apps rather than working in its own sandboxed browser alone; compared with Claude Code or OpenAI Codex, it is not coding-first.

Grok Bot is not sold on its own. On xAI's pricing page, read 9 November 2026, it comes with the Premium plan and every tier above it (Super, Ultra, Teams and Enterprise), while the Free and Plus tiers leave it out. Those plans are now sold as The X Subscription, which puts Grok, Grok Bot, Cursor and X Premium on one plan, so one weekly usage pool covers Bot work, Grok chat and Cursor's coding agents, and a heavy Bot week can eat the allowance a developer counted on. The desktop app runs on macOS and Windows, with iOS for messaging on the move, and Enterprise access goes through a sales form. Builders who want a free, self-hosted alternative should look at OpenClaw, and anyone comparing all-purpose assistants can start with ChatGPT agent mode or the agents directory.

Where the alternative wins: a Bot logs in as you, every Bot on an account shares the same machine and its saved sessions, and the vendor has not published an independent security audit of that design. Narrower jobs go to MultiOn, Jules or Gemini Spark. A code-first agent, a self-hosted one or a per-agent credential boundary points elsewhere: try Cline for code, Hermes Agent for self-hosting, or the shortlist Smart Match builds from your answers. Among the always-on agents that arrived this autumn, Claude Duet asks for approval before payments, signatures, deletions and outgoing messages, OpenAI's dots live inside ChatGPT, and Muse has a free tier aimed at sellers on Meta's apps. Also in the field: Perplexity Computer, Poke and Cue. The Grok Bot, Muse, Dots and Duet comparison sets the main four side by side. Do not buy Super or Ultra only to try a Bot: Premium already includes it, and the bigger tiers buy usage and priority access rather than extra Bot features.