Anthropic launches Claude Tag, a Slack-native agent for async team workflows

Anthropic has launched Claude Tag, a Slack-native agent designed to integrate Claude into team workflows as a persistent, delegatable teammate. The feature allows teams to grant Claude access to selected channels, tools, data, and codebases, enabling asynchronous task delegation by tagging Claude into work threads. Unlike synchronous chat-based interactions, Claude Tag is positioned for long-running, background execution, returning updates and progress without requiring continuous user attention.

The company frames Claude Tag as a significant iteration in the evolution of Claude’s user experience, describing it as a "third major redesign of LLM UI/UX" and positioning it as "Claude Code made multiplayer, async, and proactive across your whole team." Anthropic’s internal Claude Code team has reportedly used Claude Tag throughout the past year, with Anthropic employees claiming it now writes 65% of the product team’s code and merges 65% of product PRs. While these metrics likely reflect different denominators and should be interpreted cautiously, they underscore the product’s emphasis on workflow integration over standalone model performance.

Claude Tag is currently available in beta for Claude Enterprise and Team plans. Anthropic’s product lead Cat Wu described it as "our first product that is natively multi-player and proactive," and noted that there are "100s of ways" to customize the agent, with six common internal flows identified among design partners. The product requires explicit permissioning and configuration, with Anthropic publishing a guide for "agent permissions" to control access scope rather than granting blanket workspace access.

Anthropic provided a concrete example of Claude Tag’s capabilities: monitoring an A/B test with guardrails, tracking target metrics, alerting if guardrails are breached, noting mid-run corrections, and pinging the team when results reach statistical significance—all while preparing a rollout PR. This workflow highlights the agent’s ability to handle multi-step, cross-functional tasks with tool and data access. The product’s backend is implied to require robust infrastructure for identity management, permissioning, execution against external systems, task state persistence, selective context loading, and notification routing back into team workflows.

Reactions to the launch have been mixed. Supporters, including Anthropic employees and external builders, emphasize the shift from direct prompting to delegation and background execution within teams’ native communication layers. Skeptics question the UX and organizational implications, with some arguing that embedding AI so deeply into Slack may not align with existing workflows. Regardless of interpretation, the launch underscores a broader industry trend: the move from AI as a tool to AI as a persistent, collaborative participant in team processes.