Anthropic launched Claude Tag today, an "always-on" AI agent that lives directly inside Slack channels and functions as a collaborative teammate. Unlike previous Slack integrations where users DMed Claude or triggered one-off tasks, Claude Tag maintains persistent presence: it remembers context from channel conversations, can be assigned tasks via @ mentions like any team member, and proactively surfaces relevant information without being asked.
The key shift here is statefulness. Claude Tag doesn't reset between interactions. It tracks ongoing projects, recalls decisions made three days ago, and understands the implicit context of threaded discussions. For engineering teams, this means Claude can monitor deployment channels and flag potential issues before they escalate. For product teams, it can maintain running summaries of feature discussions and surface buried decisions when similar topics resurface weeks later.
Anthropic is positioning this as a shift from "AI as tool" to "AI as teammate." The agent can be added to specific channels, assigned custom roles and permissions, and integrated with external tools like GitHub, Jira, and Notion. It can draft responses, summarize long threads, extract action items, and even initiate workflows based on channel activity patterns it learns over time.
Why it matters
Slack is where work actually happens for millions of teams. By embedding a stateful agent directly into that flow, Anthropic is bypassing the adoption friction that plagues standalone AI tools. No new interface to learn, no context switching, no copy-pasting between apps. Claude Tag meets teams where they already are.
The competitive implications are significant. OpenAI has ChatGPT Enterprise and Microsoft has Copilot in Teams, but neither offers this level of persistent, channel-native presence. Google's Agentspace is still largely search-centric. Claude Tag is the first major agent product designed specifically for the messy, ongoing, multi-threaded reality of team chat.
For developers and builders, this also signals where the frontier is heading. The most valuable AI products won't be standalone chatbots. They'll be ambient, contextual, and deeply embedded in existing workflows. The race is on to own the agent layer inside the tools teams already use daily.
Also in the news
- Google unveiled Agentspace upgrades with deeper Workspace integration and multi-agent orchestration features, pushing their enterprise AI strategy forward.
- OpenAI is reportedly testing persistent agents for ChatGPT Enterprise that can maintain context across sessions, though no public launch timeline has been announced.
- Microsoft announced Copilot enhancements for Teams with better meeting summarization and action item extraction, continuing the enterprise AI arms race.
- Slack's native AI features got a quiet update this week with improved search and thread summarization, possibly a defensive move against third-party agents.