Anthropic launched Claude Tag on June 23, a persistent autonomous agent that lives inside Slack workspaces and replaces the company’s previous stateless chat integration. The product runs on Opus 4.8 and is available in beta to Claude Enterprise and Team plan customers. The old Claude-in-Slack app shuts down on August 3, giving workspace administrators 30 days to migrate.
The headline number from the launch announcement: 65% of Anthropic’s own product team code changes are now created by their internal version of Claude Tag. That figure describes current production usage at the company that builds and trains Claude, not a benchmark demo.
How Claude Tag Differs from the Old Integration
The previous Claude-in-Slack was stateless. You tagged it, got a response, and the thread ended with no memory carried forward. Claude Tag operates as a persistent identity inside the workspace. It maintains context across entire conversation threads, breaks complex requests into sequential sub-steps, and executes them without requiring follow-up prompts from users.
Three capabilities separate it from the predecessor, according to Anthropic:
Claude Tag is multiplayer. One Claude instance serves an entire channel, visible to everyone. Any team member can pick up a conversation where someone else left off, closer to a shared teammate than a personal assistant.
Claude Tag accumulates context. It follows along with channel activity and builds knowledge about the work over time. Administrators can grant it access to additional Slack channels and data sources. Users no longer need to re-explain project context from scratch each session.
Claude Tag takes initiative. With ambient mode enabled, it monitors threads proactively, flags relevant information from across connected channels, and follows up on stalled tasks without being asked. It can also schedule tasks for itself and work autonomously over hours or days.
Access Controls and Administration
Anthropic built the product around enterprise access management. System administrators define which tools, data, and channels each Claude Tag instance can reach. Separate instances for different teams (sales, engineering, support) maintain isolated memory scopes, so a model configured for sales work cannot pass context or data to one configured for engineering.
Administrators can set token spend limits at both the organization and channel level, and can view a complete log of every action Claude Tag has taken along with who requested each task. Anthropic is offering introductory launch credits to eligible Enterprise and Team organizations.
The August 3 Migration Deadline
The old Claude-in-Slack integration stops working on August 3, 2026. Teams still using the legacy integration after that date lose Claude access in Slack entirely until they complete the migration. There is no data migration required since the old integration maintained no persistent state.
Claude Tag is not available on free or Pro plans. Bind AI’s analysis notes this positions the product squarely at organizations with real collaborative workflows rather than individual productivity use cases.
Competitive Positioning
Claude Tag enters a market where Microsoft Copilot in Teams is the closest native messaging-platform competitor. The key differentiators are ambient behavior (Copilot responds to explicit prompts only, while Claude Tag monitors and surfaces information independently) and platform flexibility (Copilot requires Microsoft 365, while Claude Tag operates within existing Slack workspaces).
Anthropic framed the product as “the beginning of an evolution of Claude Code,” extending the same autonomous task execution pattern from developer terminals into team collaboration. The company says the pattern is already spreading beyond engineering internally, with teams using Claude Tag for product metrics, support tickets, and debugging.