Connect it to your Slack workspace
A workspace admin enables Claude Tag for the organisation. If you already ran the old Claude in Slack app, Anthropic gives you a 30-day window to migrate the existing setup across.
Claude Tag guide
How to use Claude Tag.
Claude Tag is Anthropic's always-on teammate inside Slack. Here is how to switch it on, talk to it, and put it to work - in plain steps, no hype.
The short version.
Claude Tag launched in June 2026 to replace Anthropic's older Claude in Slack app. Instead of a chatbot you open in a side panel, it is one shared Claude that lives in a channel, learns from the conversation, runs on the Opus 4.8 model, and in ambient mode speaks up on its own. It is in beta for Claude Team and Enterprise customers. Using it well comes down to four things: turning it on, tagging it, letting it run, and scoping what it can touch.
Turn it on.
Claude Tag is an admin-enabled app, not something each person installs. It needs a Claude Team or Enterprise plan today.
Add @Claude to the channels that need it
Invite it into a channel the same way you would add a person. There is one Claude per channel, and everyone in that channel shares it - the work it does is visible to all of them.
Scope what it can see and do
Per channel, admins choose which tools and data sources Claude can reach and what it is allowed to remember. You can also set token-spend caps for the whole organisation and for individual channels so cost stays in bounds.
Tag it, or just talk.
Once it is in a channel there are three ways to work with it.
Mention @Claude in a channel
Tag it in a message with the request. It breaks the task into stages, works through them with the tools it has access to, and replies in the thread so the whole channel can follow along.
Send it a direct message
DM Claude when you want a private, one-to-one answer that the channel does not see - useful for a quick draft or a question you would rather keep to yourself.
Reply in the thread to steer it
Follow up in its thread to correct or refine. Because there is one Claude per channel, anyone can pick the task up and the whole channel sees the exchange.
Let ambient mode do the watching.
Ambient mode is the part that makes Claude Tag feel like a teammate rather than a command line. When an admin enables it, Claude stops waiting to be called: it follows the channels it is in, flags information you might have missed from elsewhere in the organisation, and nudges threads that have stalled - a decision still open, a problem reported days ago that no one closed.
It can also work asynchronously. Hand it something larger and it will schedule tasks for itself and push a project forward over hours or days, posting back in the thread as it goes. Anthropic says roughly 65% of its own product team's code is now produced by its internal version of this.
Ambient mode is opt-in, and it is bounded by the same per-channel permissions as everything else. Claude will not surface anything from a private channel it has not been added to.
What it can actually do.
In practice, a channel's Claude is good at four kinds of work.
Answer with context
Because it learns from the channel over time, you stop re-pasting the same background. Ask about an account or a project and it already knows the recent history.
Carry a multi-step task
It plans the steps, uses connected tools to do each one, and reports the result in the thread instead of handing you a to-do list.
Leave a trail
Admins can view a log of everything @Claude has done and who requested each task, so the autonomy stays accountable.
Get more out of it.
A few habits that make Claude Tag land better.
Put it where the work happens
Add it to the channels where decisions actually get made, not a quiet test channel. It learns from what it can see, so give it the real conversation.
Be explicit about tools
It can only act through the tools an admin connected for that channel. If it cannot do something, check the channel's tool and data scope first.
Set caps before you scale
Turn on per-channel and per-organisation token caps early. Ambient, autonomous work is useful precisely because it runs without you, so put the guardrails in up front.
A lighter way to start.
Claude Tag is a strong product, and if you are already on a Claude Team or Enterprise plan and live entirely in Slack, it is well worth turning on. If you want the same kind of teammate without the plan, the beta, or the Slack-only limit, Beagle is the lighter way to start.
In Slack and Microsoft Teams
Beagle is native to both. Claude Tag is Slack-only at launch, so if any of your team works in Teams, Beagle reaches them where Claude Tag cannot.
Free to try, nothing to migrate
No Team plan, no five-seat minimum, no 30-day migration window. Start free on one seat with 1,000 credits and no card, then add 5,000 a month for a flat $49.
Drafts first, asks before acting
Beagle prepares the reply, the update, the message - and waits for your nod before anything goes out, so you stay in control up front rather than via an audit log after the fact.
Common questions.
Do I need a paid plan to use Claude Tag?
Yes. Claude Tag is in beta for Claude Team and Enterprise customers, so a workspace admin has to enable it on one of those plans. There is no free tier today.
How do I actually talk to Claude Tag?
Mention @Claude in any channel it has been added to and describe the task, or send it a direct message for a private one-to-one answer. It replies in the thread and can keep working on longer tasks in the background.
Can Claude Tag read my private channels?
No. It only operates in channels it has been added to, and admins control which tools, data and memories it can touch per channel. It will not surface anything from a private channel it is not in.
Is there a simpler alternative to set up?
Beagle is a Slack and Microsoft Teams teammate you can invite to a channel and start using free on a single seat, with nothing to migrate and no Team or Enterprise plan required.
Read next.
Want the teammate,
not the setup?
Try Beagle free for one seat - in Slack or Microsoft Teams, today.