Claude Tag: What Anthropic's New AI Teammate Inside Slack Means for Your Team

TL;DR

Claude Tag is a new way to use Claude that lives directly inside Slack. Instead of opening a separate app, anyone on your team can type @Claude in a channel, hand it a task in plain language, and Claude breaks it into steps, works through it using your connected tools, and replies in the thread — where the whole team can watch and jump in. The big shift is that this is the first version of Claude built to be multiplayer: one shared Claude that a whole channel uses together, not a private chat each person has on their own. It can also work ambiently — quietly following up on forgotten threads and keeping people posted without being asked each time. Anthropic launched it as a research preview for Claude Team and Enterprise customers on June 23, 2026. Here's what it is, why a shared AI teammate is different from everything that came before, and how to think about it for your team.

What Claude Tag Actually Is

It's Claude as a coworker that sits inside Slack, that anyone in a channel can tag into a task — and everyone can see the work as it happens.

Up to now, using Claude meant going somewhere to use it: open the Claude app, start a chat, give it context, get your answer, come back later. That works, but it's fundamentally a solo experience. You have your conversation, your colleague has theirs, and the two never meet. The work Claude does for you lives in your account, invisible to everyone else.

Claude Tag flips that. Claude becomes a participant in the place your team already talks all day — Slack. You bring it into a channel, and from then on anyone in that channel can write @Claude followed by a request, exactly the way they'd ping a colleague. Claude reads the request, figures out what needs to happen, does the work, and posts the result back in the thread. Everyone in the channel sees the question, the work, and the answer.

The simplest way to picture it: until now, Claude was a brilliant assistant who only ever worked behind a closed office door, one visitor at a time. Claude Tag moves that assistant out to a desk in the middle of the room, where the whole team can walk up, ask for something, watch it get done, and pitch in if it's heading the wrong way.

The One Word That Changes Everything: Multiplayer

Claude Tag is a single shared Claude that a whole team uses together — not a private copy each person has.

This is the part worth slowing down on, because it's the real news. Anthropic has described its earlier products — chat, Claude Code, Cowork — as essentially single-player. You use them alone. Claude Tag is the first product built to be multiplayer: there's one Claude “identity” in a channel, and everyone in that channel works with the same one.

That sounds like a small distinction. It isn't. It means a task no longer belongs to one person. Someone in marketing can ask Claude to start pulling together a campaign brief, step away for a meeting, and a colleague can pick up the same thread, see exactly where Claude got to, and nudge it forward — without anyone re-explaining the context. The work is shared the way a Slack conversation is shared, because it is a Slack conversation.

It also means the team learns together. When one person figures out a clever way to get Claude to do something useful, everyone in the channel sees it happen in real time. The good habits spread by being visible, instead of staying locked inside one person's private chat history.

How a Task Actually Flows

You tag Claude with a plain-language request; it breaks the task into stages, works through your connected tools, and reports back in the thread.

Here's the everyday rhythm of it. Someone types something like “@Claude pull last month's support tickets, group them by theme, and draft a summary for the leadership channel.” Claude doesn't just fire back a one-line answer. It breaks that into steps — find the tickets, read them, cluster the themes, write the summary — and works through them on its own, reaching into the tools and data it's been connected to. When it's done, it posts the finished summary right there in the thread.

Because it's all happening in Slack, the work is legible. You're not staring at a spinner wondering what's going on. You see Claude take the request, see it work, and see the result land where the rest of the team can read it, react to it, and act on it. If it's heading somewhere unhelpful, anyone can jump in mid-stream and correct course — “actually, only include enterprise customers” — the same way you'd redirect a colleague.

The “Ambient” Part: Claude That Follows Up Without Being Asked

Claude Tag can proactively keep people updated and chase down forgotten threads — not just respond when tagged.

Most AI tools are purely reactive: they do nothing until you prompt them. Claude Tag adds something Anthropic calls ambient behavior — the ability to act in the background without a direct request each time.

In practice, that looks like Claude noticing a thread where a question was asked and never answered, and gently following up. Or keeping a channel posted on the status of a longer task it's working through, so nobody has to ask “is that done yet?” Or surfacing a relevant piece of information from elsewhere in the organization that the people in this conversation would want to know.

The value here is the stuff that normally falls through the cracks — the follow-up nobody remembered, the loop that never got closed, the update someone forgot to send. A teammate who quietly handles those things is worth a lot, and that's the role Claude Tag is reaching for. It's a meaningful step up from “answers when asked” toward “notices what needs doing.”

The Guardrails: Tightly Scoped Access

Admins control exactly which tools, data, and memories Claude can reach in each channel — so HR's Claude can't leak into engineering's.

The obvious worry with “one shared Claude that follows conversations and connects to your tools” is: does it now know everything, and could it spill sensitive information into the wrong room? Anthropic built Claude Tag with that exact concern in mind, and the answer is that access can be very tightly scoped.

System administrators decide which tools, which information, and which memories Claude has in which channels. The Claude that works in the HR channel can be given HR's tools and context and nothing else — and it won't carry that information over to the engineering channel, or vice versa. You can think of it less as one all-knowing entity and more as the same trustworthy colleague who simply doesn't bring confidential conversations from one room into another.

For a non-technical leader, the practical takeaway is this: the multiplayer, always-present nature of Claude Tag doesn't mean a free-for-all. The controls over what it can see and do are set centrally, channel by channel, by whoever administers your Slack and Claude setup. That's the feature that makes a shared AI teammate safe to actually deploy across departments.

Why This Matters More Than It First Looks

Claude Tag meets your team where the work already happens, and turns AI from a private tool into shared infrastructure.

The quiet genius of putting Claude in Slack is that it removes the biggest barrier to AI adoption that we see over and over: people don't change where they work. You can give a team the best AI app in the world, and most of them still won't open a second window in the middle of a busy day. But they're already in Slack. Bringing Claude to them — instead of asking them to go to Claude — closes a gap that no amount of training fully closes on its own.

There are three compounding effects worth naming:

Adoption stops depending on individual willpower. When using Claude is as simple as @-mentioning it in a channel you're already in, the friction that kills most AI rollouts largely disappears. The tool is where the work is.
AI knowledge becomes a team asset, not a personal one. Because the work is visible in shared channels, the people who are good at directing Claude teach everyone else just by doing it in the open. Skill spreads instead of staying siloed.
Tasks survive handoffs. A half-finished piece of work in a channel can be picked up by anyone, because the context lives in the thread, not in one person's private session. That's how real teams actually operate — and now the AI fits that pattern instead of fighting it.

Five Concrete Ways a Business Team Uses Claude Tag

Examples that map onto work your team is already doing in Slack.

1. The shared customer-question channel

A support team has a channel where tricky customer questions land. Instead of one person researching each one, anyone tags @Claude to pull the relevant policy, draft a reply, and post it in the thread for a human to review and send. The whole team sees the answers, so the same question never gets researched twice.

2. The recurring status roundup

A project channel asks Claude to compile a weekly status — what moved, what's blocked, who owns what — and post it every Monday. With ambient behavior, Claude can chase the open items that nobody updated, instead of the project lead spending Friday afternoon nagging people.

3. The marketing brief that passes between hands

Someone kicks off a campaign brief by tagging Claude with the goal and audience. Claude starts drafting. A colleague who knows the product details jumps into the same thread, adds context, and steers Claude to refine it. The brief gets built collaboratively, with Claude doing the assembling and the team doing the steering.

4. The “catch me up” in a busy channel

A manager returns from two days off to a channel with 200 messages. They tag @Claude to summarize what happened, flag anything that needs a decision, and list what's waiting on them. Five minutes of reading replaces an hour of scrolling.

5. The operations follow-up nobody remembers

An operations channel relies on small commitments — “I'll send that over by Thursday.” Claude quietly tracks the open loops and follows up on the ones that go quiet, so the things that usually slip through simply don't.

How Claude Tag Fits With the Rest of Claude

Chat, Cowork, Claude Code, and Claude Tag each fit a different shape of work — and Tag is the team one.

If you've followed our earlier pieces, it helps to place this alongside the others, because the names blur together:

Claude chat is the one-on-one workspace — you, asking and refining, in your own conversation.
Cowork is the do-it-for-me mode — you hand off a bigger task and Claude works through it and delivers a finished result.
Claude Code is the most autonomous, capable end — building tools and working through complex, multi-step jobs.
Claude Tag is the team one — a shared Claude in Slack that a whole channel uses together, with work visible to everyone.

The clean way to hold it: the first three are about you and Claude; Claude Tag is about your team and Claude. They don't compete — many companies will use the private modes for deep individual work and Claude Tag for the collaborative, in-the-flow-of-the-day work that happens in shared channels. You don't need all of them to start, and Claude Tag is the natural entry point if your team already lives in Slack.

The Mistakes Teams Will Make With Claude Tag

A few avoidable errors that keep a shared AI teammate from delivering.

Dropping it into every channel at once. Start with one or two channels where the team is willing to experiment. A focused rollout where people actually learn to work with it beats a blanket deployment everyone ignores.
Skipping the access setup. The scoping controls aren't paperwork — they're the thing that makes a shared Claude safe across departments. Decide deliberately what each channel's Claude can see and do before you turn people loose.
Treating it as a search box. The value isn't in one-line answers; it's in handing Claude a real multi-step task and letting it work. Teams that only ask trivia questions never see what it can do.
Forgetting a human still owns the outcome. Claude drafts, compiles, and follows up — but a person should still review anything that goes to a customer or a decision-maker. Use it to do the work, not to remove the judgment.
Not telling the team it's there. A shared tool only spreads if people know it exists and see it used. A two-minute demo in the channel does more than a memo.

What Your Team Should Do This Week

Three steps to get real value from Claude Tag quickly.

1. Pick one channel and one repeating task

Choose a single channel where the team is open to trying things, and one task that comes up again and again — a weekly summary, a recurring research ask, a draft someone always has to write. That's your proving ground.

2. Set the access scope before you invite people in

Work with whoever administers your Slack and Claude to decide what that channel's Claude can reach. Get the guardrails right first, so the team can use it freely without anyone worrying about what it might touch.

3. Do the first task in the open

Tag Claude with the real task in front of the channel, let everyone watch it work, and invite people to steer it. The moment a colleague sees a shared Claude finish a task they'd have done by hand — and realizes they can do the same — the idea of an AI teammate stops being abstract.

FAQ

What is Claude Tag in one sentence?

It's Claude living inside Slack as a shared teammate that anyone in a channel can tag into a task, which it then breaks into steps, works through using your connected tools, and reports back on in the thread.

How is this different from just using the Claude app?

The Claude app is single-player — you have your own private conversation. Claude Tag is multiplayer: there's one shared Claude in a channel that the whole team uses together, with the work visible to everyone and able to be handed off between colleagues mid-task.

Do I need to be technical to use it?

No. You use it by typing @Claude in a Slack channel and describing what you want in plain language, exactly like messaging a coworker. The setup — connecting tools and scoping access — is handled centrally by an administrator, not by each user.

Could it leak sensitive information between teams?

Access is tightly scoped by administrators, channel by channel. The Claude in your HR channel can be limited to HR's tools and context and won't carry that into the engineering channel, or the other way around. What Claude can see and do is set centrally, which is what makes a shared AI teammate safe to deploy across departments.

What does “ambient” behavior mean?

It means Claude can act proactively without being prompted each time — following up on forgotten threads, keeping a channel posted on a longer task, or surfacing relevant information. It's a step beyond “answers when asked” toward “notices what needs doing.”

Who can use Claude Tag right now?

Anthropic launched it as a research preview on Slack for Claude Team and Enterprise customers, with plans to expand access over time. If your company is on one of those plans, your administrator can look into enabling it.

Does this replace the other ways we use Claude?

No — it complements them. Many teams will keep using private chat, Cowork, and Claude Code for individual deep work, and add Claude Tag for the collaborative, day-to-day work that happens in shared Slack channels. It's the team-facing front door, not a replacement for the rest.

Want help bringing a shared AI teammate into the channels your team already works in — with the right tasks, the right guardrails, and people who actually use it? The Deployed Kickstart gets your team hands-on with Claude in a single day, mapped to your real workflows. The Partner program gives you ongoing support to roll it out across the business and keep finding the next high-value use.