Anthropic launches Claude Tag, an always-on AI teammate in Slack

Anthropic launched Claude Tag, a new way to work with Claude directly inside Slack. Unlike a passive chatbot, Tag joins channels as a persistent team member with its own identity, memory, and the ability to act proactively and in 'multiplayer' fashion alongside humans. Claude Code lead Boris Cherny said it has 'totally changed how we use Claude' internally over the past few months, positioning Tag as the natural surface for agentic work where teams already collaborate.

Technically, Tag builds on the Claude Code and Claude Cowork lineage, layering Slack-native integrations, tool access, compute environments, and organization-wide memory so it can decompose and complete tasks across an org while accumulating institutional knowledge. Andrej Karpathy called it 'a new paradigm for interacting with Claude that is significantly more inline with all the other human activity org-wide,' noting the heavy under-the-hood engineering needed to make tools, integrations, memory, and security 'just work.'

The launch is strategically timed. Anthropic is reportedly carrying a ~$965B valuation ahead of a confidential IPO filing, and embedding Claude as a coworker in Slack deepens its enterprise footprint exactly where rivals like Microsoft (Copilot) and OpenAI are also pushing. It also fits the week's agent-as-coworker theme alongside Google's Gemini computer use and xAI's Grok /goal mode.

Skeptics flag the obvious tensions: an always-on agent with org-wide memory and channel access raises data-governance and security questions, and the 'proactive' framing can mean unwanted interruptions. The launch is a research preview limited to Enterprise and Team customers, so reliability and permissioning at scale remain to be proven — and it arrives the same week Anthropic suffered a notable service outage that drew complaints on r/Anthropic.