Team

A shared agent your team works with in channels and direct messages, with answers from private accounts posted only on approval and admin controls over spend, writes and data.

Web channels with a shared agent

Public and private channels where your team and its shared agent work in threads, with the agent answering every mention and joining in when it can help.

A question asked once in a channel gets one answer the whole team sees, follows up on and corrects, instead of each person asking the agent alone.

Private drafts from your own accounts

Ask the shared agent about your own email or calendar in a channel, and your own agent drafts the answer privately for you to approve.

The team gets answers that live in your inbox without the shared agent reading it in front of them, and nothing from your accounts posts unless you choose.

Ask what only a colleague knows

When only a colleague would know, the agent asks them, their own agent drafts an answer from their accounts and nothing posts unless they approve it.

Knowledge stuck in one person's inbox reaches the team without a meeting or a forwarded thread, and that person decides every word that leaves it.

Also in Team

Groups

An admin gathers members into a named group, so work can be shared with that group instead of one person at a time.

An audit log of who did what

Admins review who joined or changed roles, which accounts, keys and outside AI clients were connected and what was exported or shared, filtered by member or event.

Take all your team's work with you

One admin download takes every shared project, skill and table out as a zip, so a team can always leave with the work it built together.

AI spend per member, with caps

Admins see what the team spends on AI by day, week or month, per member, against monthly caps that stop a member or the whole team at the limit.

Choose your AI model provider

A team can move its agent to another model provider, such as Anthropic, OpenAI or DeepSeek, and keep searching its full history after the switch.

Approval before any change

An admin can make the agent ask before any tool sends or changes something, with a confirm card naming the exact call.

A second check on every change

Before the agent sends or changes anything, a separate model checks it matches what you asked and holds anything else for a person.

Per-tool permissions

An admin sets each connected tool to allowed, ask first or blocked for the whole team, and members can only make their own settings stricter.