Anthropic Launches Claude Tag, an AI Teammate in Slack

Anthropic on June 23 launched Claude Tag, a feature that drops Claude into a Slack workspace as something closer to a coworker than a chatbot. Instead of opening a private window to talk to an assistant, a team member simply tags @Claude in a channel and hands off a task — "draft the v5.0 release notes," say — and Claude breaks the request into stages, works through them on its own using whatever tools and data it has been granted, and posts the result back in the thread. It runs on Claude Opus 4.8 and ships first as a research preview for Claude Enterprise and Team customers.

Want to automate an agent like this? How to Use Claude Code Routines: Schedule an AI Agent to Run in the Cloud.

The design choice that matters is that Claude Tag is multiplayer. Every member of a channel talks to a single, shared Claude identity, so anyone can see what it has been working on and pick up a half-finished task where a colleague left off — the unit of collaboration shifts from the one-to-one conversation to the team channel. Claude builds context from the channel's history, and with permission can read across other channels to gather facts from elsewhere in the organization. An optional "ambient" mode lets it act without being asked: surfacing relevant information, nudging stalled threads, and following up on tasks everyone forgot — the same proactive direction Anthropic's product team has been signaling for months.

That ambient reach is exactly why the governance controls are front and center. Administrators decide which tools, data sources, channels and codebases each Claude can touch, and identities stay scoped to their channels — a legal-focused Claude cannot leak what it learned into an engineering channel. Organizations can cap token spend both company-wide and per channel, an acknowledgment that an always-on agent left unsupervised is also an always-on bill. Claude Tag replaces Anthropic's older "Claude in Slack" app, with a 30-day window for existing users to migrate, and the company is seeding adoption with introductory launch credits. Private DMs with personal tool access remain for individual work.

Anthropic's headline proof point is its own: it says 65% of its product team's code now ships through an internal version of Claude Tag. The launch slots into a widening agentic product line — alongside Claude Cowork, which has been pushing into law firms and enterprise back offices — and reflects a clear strategic read that the next phase of adoption is not better answers in a chat box but durable, accountable work that lives where teams already operate. For a plain-English primer on why "an AI that does the task" is different from "an AI that answers questions," see our guide to agentic AI.

The bet also lands Claude squarely in the most contested real estate in enterprise software. Slack is where work coordination happens, and an AI teammate embedded there competes directly with Microsoft's Copilot inside Teams and a crowd of agent startups. The open questions are the ones every "always-on" agent faces: whether ambient interjections feel helpful or like noise, whether teams will trust an agent to act across channels, and whether the work it ships holds up without a human reading every line. Anthropic's 65% figure is a strong internal signal — but the proof will be whether customers' own teams come to treat @Claude as a colleague rather than a tool they occasionally summon.