LumoMate
Home/Latest AI News/Product

Anthropic introduces Claude Tag, an always-on AI teammate for Slack and team channels

Anthropic launched Claude Tag, a version of Claude that lives inside Slack and team channels and is meant to pick up your company's context from the conversations it sees. Here is what changed, why team-context AI matters, and how to try it without leaking sensitive information.

What happened

On 23 June 2026, Anthropic introduced Claude Tag, a version of Claude that works inside Slack and team channels. Instead of opening a separate chat window, you bring Claude into a discussion the way you would tag a coworker, and it can read along and respond in the thread.

The part Anthropic emphasises is context. According to its announcement, Claude Tag is meant to pick up your company's working context from the team conversations it has access to, so it can answer using your projects, terminology, and ongoing decisions rather than starting cold each time. TechCrunch, which covered the launch the same day, framed it as Claude "learning your company, one Slack message at a time."

The headline change is where the assistant lives. Claude Tag is built to be present in the channels where work already happens, rather than waiting in a separate app for you to come to it.

Why it matters

For a beginner, the useful idea here is the gap between a chatbot and a teammate. A plain chatbot answers each question in isolation: you paste in the background, it replies, and the next session forgets all of it. An assistant that already carries your team's context can skip the re-explaining, which is most of what makes day-to-day AI help feel slow.

That is also what moves a tool from "answers questions" toward an AI agent that participates in real work: summarising a thread, drafting a reply, or pointing to the decision your team made last week. The closer the assistant sits to your actual conversations, the more it can do without hand-holding.

The same closeness is the reason to be careful. An assistant that can read a channel can read everything posted in that channel, including things people share casually. Convenience and exposure grow together here, so the value depends on setting it up deliberately rather than dropping it everywhere.

What to do next

  • Start in one low-stakes channel, not your most sensitive ones. Treat the first week as a trial where you can watch what the assistant sees and how it responds before widening access.
  • Check the admin controls and access scope before you expand it. Find out who in your organisation can add or remove it, which channels it can read, and how to turn it off. If you are not the admin, ask who is.
  • Read Anthropic's terms on data retention and security (linked in Sources) so you know how channel content is stored and handled. Do not assume how your conversations are used for training; confirm it from the official terms rather than guessing.
  • Keep secrets and customer data out of channels the assistant can read, unless your company's policy explicitly allows it. Passwords, API keys, and personal customer information should not sit in a channel an assistant is reading by default.
This briefing summarizes a public, dated announcement from Anthropic and a same-day TechCrunch report, and links to those primary sources rather than reporting anything new.
Monday 08:00, every week

One letter a week,
lasting understanding.

Only essays that don't get scrolled past. No ads, no tracking pixels, no external linkbait. The letter ends inside your inbox.

One-click unsubscribe. No spam.
Anthropic's Claude Tag for Slack, a beginner's guide | LumoMate