Anthropic launches Claude Tag, an always-on AI teammate for Slack that learns, monitors, and works autonomously.

Anthropic is introducing Claude Tag, a new service that embeds a persistent, context-aware AI assistant directly into Slack. Rolling out in beta for Slack for Claude Enterprise and Claude Team customers, the feature allows users to @mention Claude in channels to gain insights, assign multi-stage tasks, and receive proactive updates. Unlike previous integrations, Claude Tag maintains a shared memory across the channel, can autonomously gather facts from other channels (with permission), and works in an ambient mode that surfaces flags or follow-ups without being prompted. System administrators can scope each Claude identity to specific channels, preventing cross-contamination of context between, say, legal and engineering workspaces.

Why it matters: Claude Tag exemplifies the emerging "context-engineering moat" pattern, where persistent organizational memory becomes the defensible layer atop foundation models. Anthropic is betting that the deepest enterprise lock-in comes not from model capability alone but from the tacit knowledge an AI accumulates over time about a company's workflows, jargon, and decision history. This positions Claude Tag against similar plays from Microsoft (Copilot via Graph), Glean, and Snowflake/Databricks, which are all competing to be the enterprise's back-end intelligence layer. The feature also pushes against the current limits of AI agents by introducing ambient, proactive behavior — a step toward the always-on autonomous teammate that many enterprises have been hesitant to deploy on security grounds.

Grounded take: This is a product-level win for Anthropic's enterprise distribution strategy, building on its existing Slack integrations to deepen the stickiness of Claude within customer workflows. The key risk is whether organizations will grant the broad channel access necessary for ambient functionality to deliver its promise — and whether the trust required for that access outweighs the perceived security risks. If Claude Tag gains traction, it will validate Anthropic's thesis that context persistence, not raw intelligence, is the decisive factor in enterprise AI adoption, and it will accelerate the arms race among AI vendors for the enterprise knowledge layer.