Anthropic's Claude Tag Learns Your Company via Slack

Anthropic put Claude into Slack in a fundamentally different way on June 23. Previous Claude-in-Slack integrations worked like most chatbot add-ons: mention it, get an answer, start fresh next time. Claude Tag, launching in research preview for Enterprise and Team plan customers, does something different — it stays in the channel, reads what the team is discussing, and accumulates context over time.

The difference shows up in three specific features. The first is persistent memory: Claude Tag doesn't just remember the question you asked. It maintains awareness of what's been discussed in the channel and, when granted access, can pull information from other parts of the organization. Ask it about an ongoing project and it already knows the last few decisions the team made about it.

Shared AI identity, not a personal assistant

The second shift is how Claude Tag is presented to the team. Instead of each person having their own Claude session, the whole channel works with a single Claude identity. Everyone can see what it's worked on, continue tasks someone else started, and pick up where another conversation left off. In practice, Claude functions more like a shared team resource than a private productivity tool — closer to a shared inbox or a project management system than a personal chatbot.

Task management is the third piece. Users can assign Claude Tag a multi-step job — research, document drafting, tracking cross-team progress — and Claude breaks it into stages, uses whatever tools it has access to, and reports back in threads. It doesn't need to be reminded to follow up.

Ambient Mode takes this further: in this setting, Claude proactively joins conversations without being called. If someone asks a question it can answer from organizational context, it responds. If a task was assigned and never followed up on, it pings the thread. This will feel either very useful or slightly intrusive, depending on how you configure it.

The strategic picture

Anthropic is positioning Claude Tag explicitly as a play for what it calls organizational context — the accumulated knowledge of how a company works, what decisions have been made, what's in progress. The more Claude Tag observes, the more contextually useful it becomes over time.

Administrators set the boundaries: each Claude instance has defined access to specific channels, tools, and information. A legal-focused Claude can be prevented from sharing what it learns with engineering channels, for example.

The broader competitive picture is significant. Microsoft has been building this kind of ambient organizational knowledge into Copilot through its Graph integration — access to emails, documents, meetings. Anthropic is building the same capability inside Slack rather than through Microsoft's ecosystem. For teams where Slack is the real center of gravity, Claude Tag is a more natural fit. Claude Tag is available in research preview now; wider access isn't yet announced.