Wink Pings

Grok Bot 1.0 Has Landed in Slack—Handle Permissions Before Deleting the Bot

xAI's Grok Bot 1.0 is starting to emphasize team sharing, Slack/Teams, and meetings. Someone laid out six ways to build a bot, from one environment per project to an advisor that only says no. Beyond the demo, the permission scope and project handover called out in the replies are more worth looking at first.

xAI's official Grok Bot account has released 1.0. It is positioned as an AI teammate that can be shared, collaborate in Slack and Teams, be pulled into meetings, and learn while in use. The materials say there is no pricing yet.

The official launch tweet is here.

Suryansh Tiwari listed a build approach on X. His premise is straightforward: if he were starting to build Grok bots today, he would follow these six points.

  1. One bot per major project. Each bot has its own environment, skills, and context. Toss scattered ideas to a generic 'engineering' bot. When a project ends, delete the corresponding bot.

  2. Grok coordinates; the cloud generates code. A new cloud coding session does not understand your business at first. A Grok bot familiar with the project, customers, and metrics can do better. Let it feed information to the cloud, review PRs, and then report back to you.

  3. A content bot scans X twice a day, at 9:00 and 17:00. It finds 3 topics worth posting, with the original tweet and suggested headline attached.

  4. An 'advisor' bot whose only job is to say no. Connect it to metrics, and every new idea must go through it before spending money or time.

  5. Design first, then write code. Define the design system first, for example by connecting Paper through MCP. This helps avoid some common AI pitfalls. MCP can be understood as a protocol that lets models connect to external tools.

  6. Hand off all the miscellaneous chores. Email, follow-ups, price comparisons, refunds, comparing options—these tasks are fragmented and can easily eat up a week.

Version 1.0 expands the scope to teams: shared bots in Slack, Microsoft Teams, and conference calls can learn from everyone and make sure to-dos get done.

Suryansh Tiwari's judgment is that the race between models is becoming less critical every month. The winner is the side with full context: projects, inbox, customers, finances. He says xAI has captured this layer, but pricing is not set yet. Finally, he asks developers: Would you let Grok handle your Claude, or is it just a fancy UI wrapped around Claude Code?

That question can stay open. Two reminders in the replies are more practical. Deep says to check whether each bot has its own permission scope. Alano says dedicated project environments give the coordinator clearer boundaries for code, context, and credentials. Closing a project also requires a handover: keep the final artifacts, revoke leftover access, and then let the bot disappear.

In the demo, the bot entering the meeting room is conspicuous. What really determines whether it can touch production data, customer emails, and finances is the permission scope and the access cleanup after deletion. Deleting the bot does not mean deleting the permissions it was granted.

Some people in the replies also asked the bot to 'start a new digital business,' while others requested a breakdown of the update mapped to their business. The Grok Bot account replied to the latter type of request that it would forward it to the corresponding Grok Bot.

This kind of interaction is lively, but when it comes to implementation, look at the previous paragraph first.

发布时间: 2026-11-12 08:26