A useful team bot needs access to the instructions, files and tools its colleagues rely on. It also needs to keep individual conversations private. That is the workflow behind xAI’s Team Bots launch, announced on September 28, 2026.
Team Bots are available in public beta on Teams and Enterprise plans. A team can build a Grok Bot around a role or recurring job, share it with colleagues, and invite it into a Slack channel. xAI says each person can also speak with that bot privately. Colleagues share the bot’s configuration without putting every exchange into a group chat.
One Bot Built Around a Shared Job
An account team might need one place to track customer developments; an engineering group might need help following project decisions and bug reports. Instead of having each colleague assemble a separate bot and repeatedly supply the same background, the team can give one bot access to relevant material and work from a common foundation.
People can still use the bot individually. According to xAI, it draws on shared skills while maintaining separate conversations, context and memories for each user. Colleagues share the bot’s role and expertise, but their one-to-one exchanges are not presented as a common team transcript.
This is a distinct release from the August 11 Grok Bot beta. That launch introduced persistent bots that could work across apps, remember conversations and run jobs while a user was away. xAI already described bots coordinating with one another and participating in group chats. The September update formalizes a bot that colleagues can share around a team workflow.
Four Building Blocks Give the Bot Its Working Knowledge
xAI describes four components of a Team Bot. Each adds a capability and raises a question about what the bot should be allowed to access.
- Context supplies files, instructions and skills. A team could give a bot internal documentation, a brand guide or instructions for handling a recurring process. This material helps it understand the work; it does not guarantee correct answers.
- Plugins connect the bot to applications. xAI names Salesforce, Notion and GitHub as examples, and says plugins can be connected by each person or configured for the whole team. A team-wide connection is a different access arrangement from each user bringing their own account.
- Credentials let the bot use third-party APIs that lack a plugin. xAI says this access is secure, but the launch announcement does not spell out every permission or administrative control. The scope of a credential matters as much as the bot’s instructions.
- Memory lets the bot retain what it learns over time. For a recurring job, that could reduce the need to restate decisions or corrections. xAI also says the bot maintains separate memories for individual users, distinct from the expertise the team shares.
These components give Team Bots more than a reusable prompt. A bot can have standing access to source material and applications, then carry knowledge forward as people work with it. Teams need to decide what belongs in that shared working environment.
Private Conversations and Shared Learning Need Clear Boundaries
xAI says each person’s conversations with a Team Bot remain private, with separate per-user context and memories. In a Slack channel, the bot has its own handle, and channel members can ask questions, contribute context and see its responses.
A channel response is visible to the people there; a private exchange with the same bot is not, according to xAI’s description. The announcement does not explain the technical boundary between private material and knowledge that becomes useful to the whole team.
One of xAI’s examples shows why that boundary deserves attention. The company says its internal Data Bot remembers corrections to queries so a correction from one person can improve answers for everyone. The announcement does not specify exactly how such corrections are selected or separated from a user’s private conversation and memory.
For teams considering sensitive work, the practical questions are what the bot can read, what it can act on, and what information other people may later receive. The announcement establishes xAI’s privacy claim, but it does not provide a detailed account of retention, permission enforcement or how every kind of memory is shared.
Slack Gives the Team a Common Place to Work
In Slack, colleagues can ask the bot questions and add context where the relevant team can see the exchange. They do not have to ask separately and relay answers by hand. The bot’s responses become part of the visible discussion.
xAI describes its account-team bots posting morning briefings in Slack based on sources including company news, call records, documents and relevant threads. Its engineering example uses a project channel to follow decisions, report blockers and coordinate work across connected tools. These are xAI’s descriptions of its own workflows, not independently measured productivity results.
Colleagues can also continue to work with a shared Team Bot individually. A team can maintain common instructions and tool access without requiring every request or draft to happen in a channel. It still needs to decide which questions belong there and which belong in a private conversation.
xAI’s Examples Show Ambition, Not a Benchmark
The launch announcement describes Team Bots used by xAI’s sales, engineering, marketing and analytics teams. Its marketing bot reviews drafts against brand guidance and posts a preview for sign-off on certain website changes. Its Data Bot uses shared, read-only credentials to query approved Databricks tables, according to the company.
The engineering example makes the strongest productivity claim. xAI says a bot helped coordinate a Cursor project involving hundreds of Cloud Agents, while a five-person team shipped more than 100 pull requests a day as it built Team Bots. That figure describes a company-reported internal effort. Without an independent evaluation or comparable baseline, it does not establish how many pull requests another team should expect, how much work each contained, or what portion of the outcome the bot caused.
These examples illustrate possible configurations, not promised output. A read-only data connection shows one way to limit what a bot can do. A marketing preview followed by sign-off shows where human review can remain in the process. Neither limit is necessarily applied to every Team Bot.
Teams and Enterprise Users Can Try the Public Beta
xAI says Team Bots are available now in public beta on Teams and Enterprise plans. The feature is not described as available to every Grok subscriber or as generally available.
At the August launch, xAI listed several SuperGrok and Cursor subscriptions, including Cursor Teams tiers, for the original Grok Bot beta, while directing Enterprise users to a waitlist. Its August 26 access update expanded the personal Grok Bot rollout across SuperGrok, Cursor Pro and Cursor Teams plans, but still described an Enterprise waitlist. The September Team Bots announcement explicitly includes Enterprise in the public beta for this shared feature.
Frequently Asked Questions
3 questions
1Who Can Use xAI Team Bots?
xAI says Team Bots are in public beta on Teams and Enterprise plans. Its September 28, 2026 announcement does not list a price or explain how the Teams label maps to every Cursor Teams subscription tier named in earlier Grok Bot announcements. It also does not list SuperGrok as eligible for the shared Team Bots beta.
