One always-on assistant, one identity, summoned by your whole team with a single tag.
Anthropic just launched Claude Tag, a research-preview feature that drops an always-on @Claude into Slack channels. Anyone on the team can summon it, watch what it is working on, and pick up a task where a coworker left off. The single shared identity, not a personal bot, is the real shift.
Picture your busiest Slack channel at 9 a.m. Someone asks @Claude to pull the latest numbers. Twenty minutes later a different teammate jumps in, sees exactly what the assistant already dug up, and keeps the thread moving without re-explaining a thing. No handoff doc. No “wait, who asked it what?” The assistant just stays in the room.
What Happened
Anthropic announced Claude Tag, or @Claude, currently in research preview, and pitched it as something that could eventually take over Slackbot’s job. The company describes it as an always-on Claude with a single identity that everyone in a Slack channel can access.
The mechanics are simple on the surface. Type @Claude to summon it. See what the shared assistant is working on. Pick up conversations from where someone else left off. It is less a personal chatbot and more a teammate that never logs off.
The Backstory
Like other AI bots, @Claude learns more about the kind of work a channel does over time, so there is less and less need to explain things to it. The longer it sits in your channel, the more context it carries.
It can also reach beyond a single channel. The @Claude for a specific division or team can gather information from elsewhere in the organization, as long as it has permission to do so. That permission piece is doing a lot of quiet work here, and Anthropic clearly knows it.
The Plan
Here is where it gets ambitious. If an admin enables “ambient” behavior, @Claude can proactively update people in a channel about relevant information from other parts of the organization. It can also follow up on tasks that have not had an update in a while.
Read that again. Not just answering when asked, but nudging the team on its own. The assistant stops being a tool you reach for and starts being a colleague that taps you on the shoulder.
Claude Tag is rolling out first in research preview on Slack for Claude Enterprise and Claude Team customers. Anthropic is expected to widen availability later.
The Business Model Angle
This is a distribution play dressed up as a feature. Anthropic is not asking you to open a new app or learn a new interface. It is planting Claude inside the place your team already lives all day, then letting shared context compound until switching feels expensive.
That is the same move that made Slack itself sticky, and the same playbook behind every workflow tool that ever won. Embed where the work already happens, accumulate context that only you hold, and the integration quietly becomes a moat. The single shared identity sharpens it further. A personal bot leaves with the employee. A channel-level assistant that knows the team’s history belongs to the team, which means it belongs to whoever pays the Slack and Claude bills.
The lesson for founders is blunt. The winning AI product is rarely the smartest one in a demo. It is the one that gets closest to where decisions actually get made, and makes leaving painful.
The Risk
The ambition cuts both ways. An always-on assistant that pulls data across an organization and proactively pings people is exactly the kind of thing that makes security teams sweat. One channel with the wrong permissions and “helpful context from elsewhere” becomes “leaked information that should have stayed put.”
Anthropic’s answer is control. It says @Claude’s access to sensitive data and task-specific tools “can be very tightly controlled,” with a systems admin specifying which tools and information the assistant gets for each channel. That is reassuring on paper. In practice, governance is exactly where this gets hard, and companies are already drowning in it. We covered how firms built too many AI agents and now have to corral them, and a proactive channel assistant adds a fresh layer to that mess. The feature is only as safe as the least careful admin who configures it.
Quick Questions
What is @Claude in Slack?
It is Anthropic’s Claude Tag, an always-on assistant with a single shared identity that any member of a Slack channel can summon by typing @Claude. It is currently in research preview.
How is it different from a normal Slack chatbot?
The big difference is the shared identity. Everyone in the channel uses the same @Claude, can see what it is working on, and can pick up tasks another person started. It also learns the channel’s work over time.
Can @Claude access private company data?
Only what an admin allows. Anthropic says access to sensitive data and tools can be tightly controlled, and a systems admin sets which tools and information @Claude can use per channel.
Who can use it right now?
Claude Tag launches in research preview on Slack for Claude Enterprise and Claude Team customers, with broader availability expected later.
The Business Model Analyst Take
@Claude in Slack is not really about a smarter bot. It is about location. Anthropic is betting that the assistant living inside your team’s daily conversation, holding context nobody else has, beats a more powerful model sitting one tab away. For founders, copy the strategy, not the hype: win the spot where the work already happens, let shared context pile up, and make your product the thing a team cannot cleanly walk away from. Just remember that the same reach that makes it sticky is the same reach that keeps security teams up at night.
Based on reporting by Mariella Moon for Engadget.
