Your Slack channel probably has tasks nobody owns, threads that died mid-update, and context that gets re-explained every time someone new touches the project. That’s a coordination tax — and it compounds the faster your team moves.
On June 23, 2026, Anthropic changed how that works inside Slack. @Claude now joins a channel as a shared team member — not a private chatbot you query in isolation, but one AI the whole channel delegates to, with growing memory of what your team is actually working on. TechCrunch reported that 65% of Anthropic’s own product team’s code now flows through their internal version of it. That number tells you more than any marketing copy could.
Here’s what Claude Tag is, how to delegate tasks that actually land, when ambient mode earns its keep, and what this costs before you loop in your IT admin.
Claude Tag is a channel member, not a query box
Most AI integrations in Slack work like a vending machine. You put a prompt in, get an answer out, and the next person starts from zero. Claude Tag is different by design.
When a workspace admin adds @Claude to a channel, it joins as a shared member with its own persistent identity and growing context about that channel’s work. Tag it with a request Monday morning. A colleague picks up the thread Wednesday afternoon. Claude has been following the channel the whole time — it knows what was decided, what’s still open, and how your team talks about its projects. Nobody writes a three-paragraph briefing before asking a question.
That’s the shift worth understanding. In most team AI setups, the bottleneck isn’t model capability — it’s constant re-explaining. Claude Tag removes that loop, at least within a channel.
One thing to understand clearly: context is channel-scoped. If @Claude is in both your #marketing channel and your #engineering channel, those are treated as completely separate identities with separate memories. The information doesn’t cross. Admins define which tools, data sources, and codebases each channel’s Claude can reach — which is useful for security but means you can’t assume Claude knows your whole organization from one channel assignment.
How to actually delegate tasks to @Claude
The mechanics are close to how you already assign work to a teammate. Tag @Claude in plain language and say what you want.
“Pull last week’s signups from the database and flag anything unusual.” Or: “Review the open tickets in this channel and tell me what’s still unresolved.” Claude breaks the request into stages, works through them using whatever tools your admin connected — Notion, GitHub, Google Drive, internal APIs — and posts the output in a thread under your original message.
You don’t need to stay in Slack while it works. That’s the actual value proposition. Set the task, close the window, and the result is in the thread when you’re back. The whole channel sees it. Anyone can comment, redirect, or hand the conversation off — the work is visible and shared, not buried in someone’s private DM history with a chatbot.
A few patterns that consistently improve results:
- Specify the output format. “Write a summary” returns something very different from “write a bulleted list under 150 words I can paste into the Monday standup.”
- Name which connected tool to use when there are options. “Check the Notion project board, not just what you remember.”
- If Claude pauses to ask a clarifying question, any team member can answer. You don’t have to be the one to unblock it — that multiplayer quality is the whole point.
The last point matters more than it sounds. Because the whole channel manages the conversation, tasks don’t stall just because one person is stuck in back-to-back meetings.
Ambient mode: when to enable it and when to skip it
Beyond direct requests, Claude Tag has an optional ambient mode that admins can toggle per channel. With ambient on, @Claude monitors the channel proactively — following up on threads that went quiet, flagging relevant updates from connected tools, surfacing information before anyone asks for it.
In the right channel, this earns its keep. Think of a product ops or support channel where tickets get logged but nobody consistently has the bandwidth to chase every one. Ambient Claude flags what’s been sitting for three days without movement. Nobody has to remember to check. The channel clears faster.
In a high-volume channel where people already post quickly and carefully, ambient mode adds notifications nobody asked for. The signal-to-noise ratio flips, and it becomes the kind of thing you mute.
The practical call: leave ambient off by default. Enable it specifically on channels where follow-through is the actual problem — where things genuinely fall through the cracks because everyone assumes someone else is watching. It’s a channel-level setting, not workspace-wide, so you can tune it precisely rather than choosing all-or-nothing.
What Claude Tag handles well — and where it still needs you
From Anthropic’s documented use cases and the reporting at launch, Claude Tag performs well on:
- Data summarization and anomaly detection — pull numbers from a connected source, summarize trends, flag outliers. Strong when the data is actually in a linked tool. Weaker when it isn’t.
- Code tasks — Claude Tag shares architecture with Claude Code. The 65% internal stat points to genuine capability here, not just positioning.
- Support ticket triage — categorizing, prioritizing, and drafting responses for human review. It handles the work before the judgment call, not the judgment call itself.
- Long-horizon research — synthesizing information across documents and sources that would take hours of reading. Delegate asynchronously and come back to a working draft.
Where it still needs you: anything requiring real-world judgment about relationships, risk, or context that isn’t in the connected tools. If your data lives in a system that’s not linked to Claude’s channel access, it can’t reach it. And because its memory is channel-wide rather than per-user, think carefully before giving @Claude access to confidential information in a channel where that context could surface to colleagues who shouldn’t see it.
Async AI work also means mistakes can accumulate quietly. Build a habit of reviewing outputs before acting on them — especially for anything customer-facing, financially consequential, or going out under your name.
Who can use Claude Tag today — and what it costs
Claude Tag is in beta for Claude Team and Enterprise customers. Free and individual Pro plans aren’t included.
At the time of writing, Claude Team Standard runs $25 per seat per month ($20 per seat billed annually) for organizations with 5 to 150 seats. Team Premium — which adds Claude Code and Cowork features — is $125 per seat per month ($100 billed annually). Enterprise pricing is custom; Anthropic decoupled tokens from seats in 2026, so the base fee (roughly $20 per seat per month) doesn’t include usage — API consumption is metered separately on top. Plans have been moving fast; verify current numbers on Anthropic’s official pricing page.
If you’re currently running the older Claude in Slack integration, there’s a deadline. Claude Tag replaces it, and workspace admins have 30 days from the June 23 rollout to opt in — roughly until early August 2026. After that window, the old app is discontinued. If you’re the admin, don’t leave this for the last week.
If you’re not yet on a qualifying plan and want to try Claude before committing, you can start with Claude for free to get a feel for how it works — though Claude Tag itself requires a Team or Enterprise plan.
Frequently asked questions
Q. Does @Claude read all the old messages in my channel when it first joins?
No — Claude Tag builds context progressively as it follows the channel going forward, not from a full historical download. Think of it as a new hire who reads the room as conversations happen rather than receiving a complete transcript on day one. What it accumulates from the moment it joins is available to it; prior message history isn’t retroactively ingested.
Q. Is Claude Tag available on the free Claude plan?
No. As of the June 2026 launch, Claude Tag is available only for Claude Team and Enterprise customers. The minimum entry point is Claude Team Standard ($20 per seat per month billed annually). There’s no announced timeline for expansion to individual plans.
Q. Can I use Claude Tag outputs in commercial or client-facing work?
Generally yes under Team and Enterprise plans, but the specifics vary by plan and any additional agreements your organization has with Anthropic — and commercial use terms often differ between plan tiers and free-tier access. Before publishing AI-generated content commercially, confirm your plan’s terms at the Anthropic site and check with your organization’s legal or compliance team. Don’t assume channel AI context automatically grants commercial clearance.
Where to start
If your team is on a Claude Team or Enterprise plan and already runs in Slack, the highest-leverage first move is simple: pick one channel — the one with the most coordination friction, the most tasks falling through the cracks — and add @Claude there. Connect one or two tools that channel actually uses. Run it for two weeks before expanding.
Teams that get the most out of Claude Tag aren’t the ones who add it everywhere at once. They find one real bottleneck, solve it, and measure whether it worked. That approach also makes it easy to justify the seat cost to whoever holds the budget.
If you’re a workspace admin with the old Claude in Slack still running, check your migration timeline now. The opt-in window closes around early August 2026. Rebuilding from scratch after a lapse is avoidable, and the setup isn’t complicated once you’ve cleared the admin steps.