Anthropic launched Claude Tag, a Slack-native agent that any employee can summon by tagging @Claude in a channel. Claude Tag accepts a task, breaks it into stages, and posts the result back asynchronously — so the person who delegated the work does not have to babysit it. The release is in beta for Claude Enterprise and Claude Team customers, runs on Opus 4.8, and is governed by admin controls that scope which tools, channels, and memories the agent can touch.
Channel-level deployment is a meaningful step beyond a chatbot. Every employee in the workspace shares one Claude identity with shared context, so a teammate can pick up an in-flight task the way a human collaborator would. Anthropic disclosed that 65% of its product team's code is currently generated by an internal version of the same tool — a self-reported figure but a useful disclosure about how the company sees the maturity of the workflow.
Slack is the third major workplace surface Anthropic has targeted in 2026, after the Claude API and the Claude developer CLI. The launch overlaps with OpenAI's GPT-5.5 enterprise push, the recent company-wide Codex rollout at Samsung, and Salesforce's own Slack-integration plans. The signal is consistent: frontier vendors now view embedded workplace agents — not standalone copilots — as the next wedge into enterprise spend.
A takeaway for learners: tools that need you to copy-paste context will feel old quickly. The pattern moving forward is an agent that lives where the work already happens, inherits the conversation, and accepts multi-step delegation. If you are learning to use AI professionally, get comfortable with the shape of "hand a multi-step task to an agent with shared memory, then verify the result" — that workflow now matters more than per-prompt accuracy.