Anthropic debuts Claude Tag AI agent in Slack for team collaboration

Anthropic on Tuesday officially launched Claude Tag, a new AI agent product integrated directly into Slack, deepening the company's push into the enterprise collaboration market. Unlike previous one-on-one Claude integrations, Claude Tag operates within shared Slack channels, allowing all team members to interact with the agent simultaneously, view its progress, and pick up conversations where others left off. It also features continuous context accumulation — the agent learns from ongoing channel activity, Slack data sources (excluding private channels), and retains knowledge over time, eliminating the need to repeatedly re-explain requirements. Users can assign tasks and let Claude Tag work autonomously over hours or days, even scheduling its own tasks. Anthropic has been using internal versions of this technology to dramatically boost productivity; about 65% of the code for their product team is now generated by Claude Tag agents assigned in parallel.

However, the launch was complicated by regulatory hurdles. Less than two weeks before Claude Tag's release, Anthropic suspended access to its most advanced models, Fable 5 and Mythos 5, due to a Trump administration ban aimed at preventing foreign nationals from acquiring the technology. Cat Wu, product lead for Claude Code, confirmed that Anthropic originally planned to use both Fable 5 and Opus 4.8 as the underlying models for Claude Tag. She described Fable 5 as the optimal model — significantly better at coding, executing tasks with minimal user assistance, and deciding when to step into conversations. The forced pivot to only Opus 4.8 may limit the agent's peak performance, though Anthropic believes the product still delivers substantial enterprise value.