Claude Tag turns Slack channels into shared AI-agent workspaces

The next serious interface for AI agents may be the team channel, not the personal chat window. Anthropic introduced Claude Tag on June 23, 2026, making it available in beta to Claude Enterprise and Team customers on Slack. Users can mention @Claude inside a channel, then delegate work within the tools, data, and codebases an administrator has explicitly approved.

The product matters because it changes the collaboration model. Claude Tag is not framed as a private assistant that answers one person at a time. Anthropic describes it as a shared Claude inside a channel: team members can see what it is doing, continue a thread from where someone else left off, and benefit from context accumulated in the places where work already happens. That makes the AI agent closer to a team resource than a solo productivity feature.

Anthropic is also making a pointed internal-use claim. The company says 65% of its product team’s code is created by its internal version of Claude Tag. It also says teams use the same pattern to investigate product metrics, work through support tickets, and find root causes for difficult bugs. That number is not an independent benchmark, but it is a useful signal about where Anthropic thinks Claude Code evolves: from an individual coding tool into an organization-level delegation system.

The access model is central. Administrators decide which channels, tools, and information Claude can use. They can create separate Claude identities for separate functions, keeping sales context away from engineering context and preventing memories from crossing boundaries. Anthropic also lists organization and channel token-spend limits, action logs, and request attribution, which are the controls enterprises usually need before letting an agent act inside shared systems.

Claude Tag replaces the existing Claude in Slack app, with a 30-day opt-in migration path for administrators. It runs on Opus 4.8 and can work asynchronously, including scheduling tasks over hours or days. The open question is whether channel memory and ambient updates reduce coordination cost or create another stream of work to govern. Either way, the direction is clear: enterprise AI assistants are moving from answering prompts to occupying the same collaborative spaces where teams assign, audit, and finish work.