Anthropic Just Put Claude Inside Your Slack. Now What?

Claude is no longer a tab you open. It's a colleague you tag.

On 23 June 2026, Anthropic launched Claude Tag, a new product that places Claude directly inside Slack as a persistent, shared team member. Available immediately in beta for Claude Enterprise and Team customers, it lets anyone in a channel tag @Claude with a task and delegate work to it while they focus on something else.

This isn't an incremental update to the existing Claude in Slack integration. It's a structural rethink of how AI fits into how teams actually operate, and if you're a founder, CMO, or growth leader, you should be paying attention to the mechanism, not just the headline.

What exactly changed, and when

The existing Claude in Slack app is being replaced by Claude Tag. Administrators of Enterprise and Team accounts have a 30-day opt-in window from 23 June 2026 to migrate. Anthropic is issuing a launch credit to eligible organisations so that teams can trial it at scale before committing spend.

Claude Tag runs on Opus 4.8. It works today, not in preview for a future quarter.

The four mechanics that actually matter

Four design decisions separate Claude Tag from anything that came before it, and each one has a direct implication for how your team should think about deploying it.

1. It's multiplayer by default

One Claude instance lives inside a channel and is visible to everyone in it. Anyone can pick up a thread mid-task. That's a fundamentally different model from private AI chat, where context dies with the conversation. Think of it less like a chatbot and more like a shared inbox that can execute.

2. It builds persistent memory

Claude accumulates context from the channels it's in, so colleagues don't need to re-brief it from scratch on every request. It can also be granted permission to learn from other channels and data sources. Critically, memory stays scoped to the channels administrators define, so a Claude instance set up for sales work won't bleed context into an engineering channel or expose data it shouldn't.

3. It works asynchronously and can schedule its own tasks

Set Claude a task, walk away, and it will plan and complete work over hours or days before reporting back in a Slack thread. At Anthropic internally, teams now delegate to multiple Claude instances running in parallel. That's a workflow change worth modelling for your own team.

4. It takes initiative when you want it to

With "ambient" behaviour enabled, Claude proactively surfaces information it thinks you need, flags threads that have gone quiet, and connects dots across channels and connected tools without being asked. You can turn this off. But the option to have a standing brief running in the background is genuinely new.

The internal data point worth noting

Anthropic states that 65% of its product team's code is now created by its internal version of Claude Tag. We'd treat any single-company figure cautiously as a benchmark, but the directionality matters: the pattern is spreading beyond engineering into support, product metrics, and root-cause analysis. That's the shape of adoption to watch, not just the raw number.

Claude Tag vs. Previous Claude in Slack: Key Differences

Capability	Claude in Slack (previous)	Claude Tag (from 23 June 2026)
Channel presence	Single-user sessions	Shared, multiplayer instance per channel
Memory	No persistent context	Builds context across channel history and connected sources
Async execution	Synchronous only	Can schedule and pursue tasks over hours or days
Proactive behaviour	Responds only when asked	Can flag relevant updates and follow up on stalled threads
Admin controls	Limited	Channel-scoped permissions, spend limits, full activity log
Availability	Generally available	Beta for Enterprise and Team customers from 23 June 2026

What this means for your pipeline and your team

The practical question for a growth leader isn't "is this interesting?" It's "where does this actually change our output?" We think there are three places to look first: async research and competitive monitoring, support and success ticket triage, and internal knowledge retrieval that currently eats senior time.

The access controls matter here. If Claude can be provisioned with scoped channel identities, a sales-facing Claude can be briefed on CRM data and deal context without touching engineering infrastructure. That's the kind of separation that makes enterprise adoption viable, not just a proof of concept.

What to do right now: if you're an Enterprise or Team customer, assign an administrator to begin the migration within the 30-day window, use the launch credit to run a pilot in one high-traffic channel, and define your permission scopes before you go broad. Start narrow, build channel-specific context deliberately, and measure task completion quality before expanding ambient behaviour.

We're watching how Claude Tag interacts with AI-driven discovery more broadly, particularly as AI search and generative engines increasingly surface content that originates from team-level AI workflows. If you want to understand how your content and brand show up inside AI systems, our thinking on GEO strategy for B2B SaaS teams is a useful place to start alongside this shift.

The era of AI as a private assistant is giving way to AI as a shared colleague. Claude Tag is the clearest signal yet that Anthropic intends to win that transition at the team layer, not just the individual one. The firms that think carefully about deployment and governance now will move faster than those who treat this as a feature to try later.