Anthropic lets teams tag Claude like a coworker in Slack

Most workplace bots feel like goldfish: the moment you close the chat window, they forget everything that just happened.

Anthropic wants to fix that by turning its AI into a permanent seat in your company group chat. On June 23, the company rolled out Claude Tag, a Slack-native workspace tool powered by Opus 4.8 that lets teams delegate tasks directly in channels.

It is available in beta for Claude Enterprise and Team customers, and Anthropic plans to replace the legacy Claude in Slack app over the next 30 days.

An AI coworker that shares the room

Instead of treating AI like a private 1-on-1 chat, Claude Tag behaves like an actual coworker sitting in the channel.

Anyone in the room can tag @Claude with a request. The model breaks the project into logical steps, uses whatever tools or codebases admins have hooked up, and posts the finished work right back into the Slack thread. Because it operates publicly, other team members can see what it did and pick up the thread without re-explaining the context.

It also works asynchronously over hours or days, scheduling its own follow-ups. If teams turn on ambient mode, Claude will even chime in proactively to flag forgotten threads or surface relevant updates from other permitted channels.

Anthropic says this workflow is already running deep inside its own walls. An internal version of Claude Tag now writes 65% of its product team's code, while non-engineers use it to pull product metrics and triage support tickets.

To prevent data leaks between departments, administrators can sandbox Claude into distinct identities. A Claude assigned to sales will not leak memories to engineering, and monthly token spend caps can be enforced per channel or across the entire company.

Why it matters

We think this is Anthropic's sharpest enterprise move yet, shifting Claude from an individual productivity toy into a multiplayer asset.

When employees use standalone chatbots, prompts and context remain trapped in personal silos. Moving those interactions into shared channels creates institutional memory where everyone can inspect the work, reuse prompts, and learn from past prompts.

Still, giving an autonomous agent a public voice comes with real friction. An AI that proactively chimes in risks becoming high-volume notification spam if its judgment is off. Teams will also need to keep a tight leash on usage limits so background channel scraping does not trigger surprise cloud bills.

More broadly, the enterprise battleground has shifted toward owning workplace context. Microsoft has its Graph, Glean indexes enterprise search, and Snowflake is pitching back-end organizational memory. By anchoring Claude directly inside team chat, Anthropic is trying to win that context layer right where daily decisions are made.