What Is Claude Tag? Anthropic Put Claude in the Group Chat
One Claude per Slack channel, with a memory the whole team shares. Claude Tag is Anthropic's answer to AI's single-player problem — and the memory is the whole story.
- Claude Tag launched June 23, 2026, in beta for Claude Enterprise and Team customers, starting on Slack — one Claude identity per channel that anyone can tag.
- It keeps persistent channel memory and, with admin permission, can learn from other Slack channels and data sources — but it never accesses private channels.
- Anthropic says 65% of its product team's code is now created by its internal version of Claude Tag, per the launch announcement.
- Admins scope everything: which tools and data Claude sees per channel, token spend limits, and activity logs.
Claude Tag is Anthropic's multiplayer AI agent for Slack, launched in beta June 23, 2026 for Claude Enterprise and Team plans. One Claude identity lives in each channel with persistent memory: anyone can tag @Claude to delegate work, see what it's doing, and pick up where a teammate left off. Admins control channel access, tools, and token spend.
What is Claude Tag? It’s Anthropic dropping Claude straight into Slack — one shared AI identity per channel, launched in beta on June 23, 2026 for Claude Enterprise and Team customers. Anyone in the channel tags @Claude, hands it work, watches what it’s doing, and picks up exactly where the last person left off. The memory belongs to the channel, not to somebody’s private chat window… and that detail is the entire product.
I haven’t touched Tag yet — it’s Enterprise beta and I’m not on the list. But I run Claude Cowork at iExcel every single day. It digests my Google Meet recordings through Grain into reports before I’ve had coffee, and my whole “Let Me Cook” workflow runs through it. I’ve said it publicly: Claude Cowork may be one of the most powerful internal business tools, hands down. Tag is that idea taken multiplayer.
What is Claude Tag and what does it actually do?
Claude Tag is a multiplayer AI agent that lives inside Slack channels. Each channel gets one Claude identity with its own memory. Any member can tag @Claude to delegate a task, see what it’s working on, and pick up a thread a teammate started. It launched in beta June 23, 2026, for Claude Enterprise and Team plans.
Peep the timing… June 23rd was day eleven of the export-control blackout, while Fable 5 was still dark. Anthropic shipped a whole new product line mid-crisis.
And the receipt that matters, verbatim from Anthropic’s announcement: “Today, 65% of our product team’s code is created by our internal version of Claude Tag.” They dogfooded this into two-thirds of their own product code before showing it to us.
How does Claude Tag’s memory work?
Claude Tag keeps persistent memory per Slack channel. It remembers relevant information from the channels it’s in, and — with admin permission — can learn from other Slack channels and connected data sources. It never touches private channels. An optional ambient mode lets it proactively flag relevant information and follow up on unresolved threads or tasks.
This is where I sit up. ByteDance’s EdgeBench ran agents through 134 real-world tasks, iterating 12+ hours per task, selected runs past 72. Across roughly 38,000 hours of interaction, performance followed a log-sigmoid scaling law (R² = 0.998): competence climbs with time in the environment, then plateaus.
Translation: an agent that stays in your environment gets better in a way a fresh chat window never will. Memory isn’t the nice-to-have. It’s the engine.
Who can use Claude Tag right now?
As of early July 2026, Claude Tag is in beta for Claude Enterprise and Team customers only, starting on Slack, with Anthropic planning expansion to other collaboration platforms. System administrators specify which tools and information Claude can access per channel, set token spend limits, and get activity logs of what it did.
The admin story is real. Memories stay scoped to defined channels. Spend caps. Logs. Whoever wrote that spec got burned by an unbounded API bill once… I respect it.
Now compare the price of the other approach. On July 2nd, Microsoft announced Frontier Company — a $2.5 billion investment embedding 6,000 industry and engineering experts at customer sites. Microsoft is embedding humans. Anthropic is embedding a model in your channels. Same bet on context, wildly different cost structure.
Why does a shared AI memory matter for teams?
Because AI at work has been single-player. Every insight Claude gave one teammate lived and died inside that teammate’s private chat history. Claude Tag moves the memory to the Slack channel, so team context compounds instead of resetting — the practical difference between a chatbot and an actual AI agent working alongside a team.
I watch this at iExcel constantly. Somebody solves a problem with AI, the solution evaporates into their DMs, and three weeks later somebody else pays to solve it again. That’s not a tooling gap. That’s an amnesia tax.
One brain per channel. Everybody can read it. Everybody can add to it. That’s the play, and it’s why Tag sits at the top of my watch list for the 2026 agent stack.
Your team’s AI shouldn’t reset to zero every Monday.
#TheAIMogul
Bottom lineClaude Tag is the first serious answer to AI's single-player problem — the memory lives in the channel, not in one person's chat history. If your team runs on Slack, this is the launch to watch; if you're not on Enterprise or Team, you're waiting.