Claude Tag: a virtual coworker that lives in your Slack
In brief
Anthropic launched Claude Tag on June 23, 2026 — a research preview that puts Claude inside Slack as a shared team member. You tag @Claude into a channel, hand it a task, and it works through the steps on its own, posting progress where the whole team can see and redirect it. Here's what it does, how it differs from the Claude Slack connector you already know, and who can use it.
Contents
On June 23, 2026, Anthropic launched Claude Tag, a research preview that puts Claude inside Slack as a shared team member rather than a personal chatbot. You tag @Claude into a channel, give it a task, and it breaks the task into steps and works through them on its own — posting progress in the channel as it goes, so anyone on the team can watch, jump in, or send it in a different direction.
The framing Anthropic uses is "a coworker you tag in." That is a useful way to read the difference from what came before.
What's actually new here
Claude has been in Slack for over a year through the Slack connector — but that was Claude answering your questions in a DM or thread, tied to your account and your view of the data. Claude Tag is different in three ways:
- It's shared, not personal. Claude Tag belongs to the channel, not to one person. A teammate can hand off a half-finished task to Claude, and someone else can pick up where Claude left off. The work happens in the open.
- It works in stages, not single replies. You give it an outcome ("pull the numbers from last week's launch and draft the retro doc"), and it plans the steps and executes them, reporting back as each stage completes — closer to delegating to a person than asking a question.
- It learns your company over time. It picks up context from the channels it's in — your projects, your terminology, who owns what — so you stop re-explaining the same background on every request.
How it works in a channel
The basic loop:
- Tag it in. Mention
@Claudein a channel and describe what you want done. - It scopes the work. Claude restates the task, breaks it into stages, and starts.
- It works in the open. Progress shows up as messages in the channel. You can correct it mid-task, add a constraint, or stop it.
- It hands back. When it's done, the result lands in the channel where the whole team — not just the person who asked — can use it.
Because the work is visible, Claude Tag is also a hand-off mechanism: one person starts a task, goes offline, and a colleague continues steering Claude without losing the thread.
Three capabilities worth calling out
Ambient follow-ups. Claude Tag proactively keeps people informed and follows up on tasks that were dropped — the "you said you'd send this, want me to draft it?" behavior, but coming from the assistant instead of a manager.
Channel-scoped access. Admins control which tools, data, and memories Claude can reach per channel. That means the HR channel's Claude can't see engineering's codebase, and vice versa. This is the control that makes a shared, always-present agent safe to put in sensitive channels.
Tool and codebase connections. Claude Tag connects to the same tools and data sources you'd wire to any Claude agent, including code. As an early data point, Anthropic says that inside its own product team Claude Tag is now approving roughly 65% of submitted code changes — a sign the company is using it on real engineering work, not just demos.
Who can use it, and what it costs
Claude Tag is a research preview, available now to Claude Enterprise and Claude Team customers, with access expanding later. There's no separate price announced for the preview — it's part of the Enterprise/Team offering. As with any research preview, expect rough edges and behavior changes while it's in beta.
If you're on Pro or Max, this isn't available to you yet. The Slack connector you already have is the closest thing — personal, per-account Claude in Slack — and remains the right tool for individual use.
How to think about it if you run AI for a team
Claude Tag is the clearest expression yet of the shift from Claude as a tool people open to Claude as a participant in the room. That changes a few things you own as the person responsible for AI at your company:
- Channel hygiene becomes access control. Since access is scoped per channel, the question "which channels is Claude in, and what can it see there?" is now a real security boundary. Treat it like adding a member to a private channel.
- The work is auditable by default. Because Claude Tag posts in-channel, you get a visible trail of what it did and who redirected it — useful for trust, and for spotting where it's getting tasks wrong.
- Adoption looks different. A personal chatbot succeeds when individuals open it. A shared channel agent succeeds when teams hand work to it. Watch whether tasks actually get delegated, not just whether people say hi to it.
Related reading
- What is an AI Agent Manager (a.k.a. Agent Operator)? — the role that owns tools like this
- Claude + Slack for teams — the personal Slack connector, and how it differs
- What to share with Claude — the data-scoping framework, now a per-channel decision
- Ask Your Org — the other shared, org-aware Claude surface
Source: Anthropic launches Claude Tag, a tool that works like a virtual employee within Slack, Fortune, June 23, 2026. Claude Tag is a research preview for Claude Enterprise and Team.