Anthropic launches Claude Tag — an always-on AI teammate that lives in your Slack channels

What was launched

Per Anthropic’s announcement and corroborating coverage (Neowin, The Next Web, Benzinga, TestingCatalog):

• What it is: Claude as a member of your Slack channels. @-mention Claude to hand it a task while you work on something else.
• Multiplayer by design: one Claude per channel, shared by everyone in it — not a private 1:1 chat. It interacts with the whole channel.
• Context that compounds: Claude “learns over time, building more context about the work as it follows along with its channel.”
• Ambient mode (optional): when enabled, Claude “proactively keeps you updated about whatever it thinks you might need to know,” flagging relevant information from its channels and connected tools.
• Model: runs on Claude Opus 4.8.
• Availability: beta, Claude Enterprise and Team customers, starting on Slack.
• Controls: admins scope which tools and data Claude can touch per channel, and can create separate Claude identities for different uses.
• Anthropic’s proof point: 65% of its product team’s code is created by its internal version of Claude Tag.

Why this matters

Three reads.

1. It’s a category shift: from “tool you open” to “teammate that’s already there.” Until now, using Claude meant context-switching — open a tab, paste the relevant material, ask, copy the answer back. Claude Tag inverts that: the AI lives where the work happens and already has the context. That’s the same bet behind Gemini’s deep Workspace integration and Microsoft’s Copilot-in-Teams, but Anthropic is doing it as a shared, multiplayer presence rather than a personal assistant. The unit of AI is shifting from “my chatbot” to “our teammate,” and that’s a meaningfully different product philosophy.

2. “Ambient” proactivity is the genuinely new — and genuinely double-edged — idea. Most workplace AI waits to be asked. Claude Tag’s ambient mode acts unprompted, surfacing what it judges relevant. Done well, that’s the difference between a tool and a colleague who notices things. Done poorly, it’s notification noise or, worse, an AI surfacing the wrong information to the wrong channel. Anthropic clearly knows the risk — hence the heavy emphasis on admin-scoped access and per-channel identities. This is the feature to watch: proactive AI in shared spaces is where the value and the failure modes both live.

3. It deepens Anthropic’s enterprise push at a delicate moment. Claude Tag follows the Claude Partner Network, the KPMG alliance, and a string of regulated-industry deals — all aimed at making Claude the default work AI. The timing is pointed: Anthropic is shipping a flagship enterprise product the same month its most capable model, Fable 5, is suspended under a US export-control order, and as it heads toward a public listing. Running Claude Tag on the unaffected Opus 4.8 lets Anthropic keep advancing the enterprise story regardless of the Fable 5 standoff.

How it compares for teams choosing AI tools

If you’re evaluating AI for a team, Claude Tag changes the comparison in a specific way:

• vs. opening Claude or ChatGPT in a browser: Claude Tag wins on friction and shared context, loses nothing on model quality (it’s Opus 4.8). For teams already on Claude Team/Enterprise, it’s a clear upgrade to how the seat gets used.
• vs. Slack’s own AI and other Slack assistants: the differentiators are the underlying model (Opus 4.8 is a frontier-tier model) and the multiplayer-plus-ambient design. Slack-native AI is more deeply wired into Slack’s data model; Claude Tag bets that model quality and proactivity matter more.
• vs. dedicated agent platforms (the kind covered in best AI agents): Claude Tag is a collaboration agent, not a customer-service or sales agent like Sierra or Decagon. Different job. It’s closest in spirit to the productivity tools in our best AI productivity guide.
• vs. Claude Code: complementary. Claude Code is the terminal/IDE coding agent; Claude Tag is the team-coordination teammate. The “65% of code” stat blurs them, but they solve different problems.

The honest caveats

It’s beta, and Slack-only at launch. “Available today in beta” for Team/Enterprise means rough edges and a single surface (Slack). Teams on Teams/Discord/other tools get nothing yet, and individual Pro/Max users are excluded entirely.

The 65% figure is Anthropic’s internal metric. A company that builds frontier models, with bespoke internal tooling and an engineering culture optimized around its own AI, reporting that 65% of its product team’s code comes from Claude Tag is impressive — but it is not a benchmark an average team should expect to replicate out of the box. Treat it as aspirational signal, not a promise.

Ambient AI in shared channels is a governance question, not just a feature. Proactive surfacing across channels and connected tools is exactly where data-exposure mistakes happen. The admin controls are necessary, but they put the burden on teams to configure scope correctly. Roll it out on low-sensitivity channels first.

No independent productivity data yet. Every claim here is from launch day — Anthropic’s announcement plus reporting on it. Real-world impact on team output won’t be measurable for weeks.

What it changes for Pick Right readers

If your team is already on Claude Team or Enterprise, turn Claude Tag on in a non-sensitive channel and see whether the shared-context, delegate-by-mention model actually reduces busywork — that’s the test that matters, and the beta is free to existing seats.

If you’re an individual user, nothing changes yet; this is a team product. The Claude review recommendation on Opus 4.8 stands.

If you’re comparing AI for a team purchase, Claude Tag is a real point in Anthropic’s favor for Slack-centric organizations — weigh it against Gemini’s Workspace depth and Microsoft Copilot’s Teams integration based on where your team actually works.

For more, see the Claude review, the Claude Code review, the Claude vs ChatGPT comparison, the best AI agents guide, the best AI productivity tools guide, the Claude Partner Network coverage, and — for the contrasting backdrop — the Fable 5 government shutdown.

Questions or corrections? Email Pick Right. Want the full list? See all news.