Anthropic’s new Claude Tag update lets its Slack agent read the full conversation — and jump in unprompted

Anthropic has made significant strides in enterprise AI with its latest update to Claude Tag, the AI agent embedded within Slack channels. According to Scott White, Anthropic's head of product for enterprise, this update represents a pivotal shift towards "multiplayer AI," a move away from individual user interactions towards AI agents that operate across teams.

By enabling Claude to read context across entire conversations, the update is expected to improve its ability to decide when to proactively engage in conversations, rather than waiting for prompts. White describes Claude as evolving from a personal assistant to a company-wide chief of staff, capable of handling projects or goals that require multi-person collaboration and judgment calls.

Three key technical advancements underpin this shift: Anthropic's Model Context Protocol (MCP), which allows Claude governed access to enterprise data systems; a threshold of model intelligence that makes proactive AI agents viable; and the strategic placement of Claude within collaboration tools like Slack. The update no longer employs a classifier to determine Claude's responses, instead allowing it to read the full context of the channel and choose from four actions.

For instance, when two engineers are working on the same bug, Claude can open a thread to investigate the issue, demonstrating the potential for AI to enhance team productivity. Anthropic emphasizes the importance of restraint, with Claude going dormant in channels where it has little to contribute. While skeptics may question the tangible benefits for teams, White points to the transformation of his own workflow within Anthropic, where AI now handles more complex analyses, reducing the need for manual data work.