Claude Tag: Anthropic Put an AI Coworker in Slack. Here’s What It Can See.

Last updated: July 16, 2026

I use Claude every single day, so when Anthropic said it was turning Claude loose inside Slack as an actual teammate, I paid attention. Not a chatbot in a sidebar. A coworker with a name you can tag, a memory of what your team is working on, and the ability to go do things while you’re in a meeting.

If your job runs on Slack, this is worth ten minutes of your attention, because some version of it is probably coming to your workplace. Here’s what Claude Tag actually does, who can get it, and the part I’d want to know as an employee…….what it can and can’t see.

What Is Claude Tag?

Claude Tag is Anthropic’s AI agent that joins Slack as a team member. Admins give it access to specific channels and tools, and then anyone in those channels can tag @Claude and hand it a task, the same way you’d loop in a coworker.

Anthropic announced Claude Tag in late June, and it’s now rolling out as the replacement for the old Claude app for Slack. The company isn’t shy about how much it leans on this internally. Anthropic claims 65% of its own product team’s code is created by their in-house version of Claude Tag, and they use it well beyond engineering. They use it for chasing down metrics, working support tickets, and digging into bugs.

Slack’s own framing, via Engadget, is that Slackbot stays the built-in helper while Claude Tag is the “delegatable AI teammate.”

What Happens When You Tag @Claude

You tag @Claude with a request in plain language, and it breaks the task into stages, works through them with whatever tools it’s been given, and posts the finished work in a Slack thread. You go do something else while it works.

Three things make this different from every chatbot you’ve used:

• It’s multiplayer. There’s one Claude per channel, shared by everyone. Your coworker can see what it’s doing and pick up the conversation where you left off, the way you’d hand off with a real teammate.
• It remembers. Claude follows along with its channels and builds context over time, so nobody has to re-explain the project from scratch every Monday. With permission, it can also learn from other channels and data sources.
• It works while you don’t. Hand it a task and check the thread later. Anthropic built it on the same foundation as Claude Code, its coding agent, so the “break it down, then grind through it” behavior is the same one I lean on daily. If you’re curious what that feels like, I’ve written about using Claude Code without writing any code.

Ambient Mode: The Coworker That Never Logs Off

With “ambient” behavior turned on, Claude stops waiting to be tagged. It proactively flags information it thinks you need from across its channels and tools, and it follows up on threads and tasks that went quiet without getting resolved.

Every workplace has a graveyard of Slack threads that ended with “I’ll look into it” and were never seen again. An AI that notices the dropped ball and nudges the thread back to life is genuinely useful, and also a little unsettling, because it means something is always reading along.

Whether that lands as helpful or creepy will come down to how your company configures it, which brings us to the part that matters most.

Who Gets Claude Tag, and What It Can See

Claude Tag is in beta for Claude Team and Enterprise customers, so your company has to turn it on. Admins decide which channels it joins, which tools and data it touches, and how much it can spend, and every task it does is logged along with who requested it.

The honest catch first: you can’t get this on a personal Claude plan. If your workplace doesn’t pay for Claude Team or Enterprise, tagging @Claude isn’t happening for you yet. This is a workplace product, and the rollout runs through your IT department, not the app store.

One thing to know is Anthropic scopes it hard. Each Claude identity’s memories stay locked to the channels admins define, so the Claude set up for sales can’t leak what it knows to engineering, and vice versa. It doesn’t report from private channels. Also, there’s a full log of everything @Claude has done and who asked for it, which cuts both ways. It’s good accountability for the AI, and a record of your requests too.

Worth remembering who’s building this, too. Anthropic is the company that walked away from $200 million rather than loosen its rules on surveillance, which buys it more benefit of the doubt on workplace privacy than most. Still, if this shows up at your job, it’s fair to ask your admin two questions. The first is which channels is it in, and the second is ambient mode on.

If the delegation idea appeals to you but your office isn’t a Claude shop, the same pattern exists in tools you can set up yourself. I’ve covered the AI agent workflows worth automating first if you want to start smaller.

The office AI stopped being a chatbot in a sidebar this summer. It got a name, a memory, and a task list. However your workplace handles that, better to understand it before it shows up in your channels.

Related reading: 5 ways to use Claude Code without writing code | AI agent workflows worth automating first | New to AI? Start here