What Claude Tag Is and Why It Replaces the Old Claude App
Claude Tag is an AI workspace agent for Slack that acts as a shared virtual coworker, learning from channel conversations, remembering context over time, and automating recurring work so teams can delegate tasks instead of micromanaging a chatbot. Anthropic is replacing its older Claude in Slack connector with Claude Tag, a new agentic AI coworker that joins workspaces as a dedicated user account. Available in beta for Claude Enterprise and Team customers, Claude Tag works with Anthropic’s Opus 4.8 model and will become the default Slack experience after administrators get a 30‑day opt‑in window before the previous app is switched off on August 3, 2026. The shift matters because Claude Tag is no longer a one‑off helper in a private DM. It is a persistent presence that shares memory within channels, interacts with whole teams, and supports more complex Slack automation than the earlier, reactive integration.
From Chatbot to Agentic AI Coworker Inside Slack
Claude Tag’s core change is its move from a single-user assistant to an always-on agent shared by the entire channel. Once added to selected channels, it can read thread context, break large requests into steps, call connected tools, and post results back in-line. Anthropic frames this as “more like interacting collaboratively with a teammate” than chatting with a disposable bot. The AI workspace agent keeps a channel-scoped memory, so it can pick up where the last person left off and avoid repeated explanations. According to Anthropic, 65 percent of its product team’s code is written by Tag, underscoring how far the company is pushing agentic automation. Teams can summon @Claude for tasks like chasing product metrics, preparing call briefings, or triaging support tickets, while Slack users see a shared history of what the agent has done and can refine or reuse its work.
Proactive, Ambient Behavior and Persistent Slack Automation
Unlike typical Slack bots that respond only when pinged, Claude Tag can operate in an ambient mode where it quietly monitors permitted channels and tools. With ambient behavior enabled, it can nudge users to follow up on stale threads, flag unresolved issues, and proactively message when it spots something that appears unfinished. It can schedule tasks for itself and work over hours or days, turning Slack into an ongoing execution environment rather than a simple chat surface. Because Claude Tag builds context from all allowed channels, a support channel agent can draw on engineering or product conversations when diagnosing bugs, or pull metrics from analytics tools to enrich updates. This persistent, cross-channel awareness makes it a more capable AI workspace agent, but also demands careful admin configuration so its always-on presence feels helpful rather than intrusive in busy workspaces.
How Teams Use Claude Tag Day to Day
In daily use, Claude Tag centers on the @Claude mention: anyone in a channel can delegate work by tagging the agent in a thread. From there, it can draft documents, summarize long discussions, transform notes into project plans, or turn bug reports into draft pull requests when connected to code repositories. It can also monitor backlogs, triage tickets, and pull data on demand, effectively acting as an agentic AI coworker that spans roles. Because Claude Tag is shared, teammates can see each interaction, refine prompts, and reuse outputs, which supports more transparent Slack automation than scattered private chats. Direct messages remain available for private queries or sensitive tasks, with the same tools and web search (where enabled) accessible from a side panel. Over time, Tag’s channel memory reduces context-setting overhead, making routine chores—status updates, follow-ups, report prep—far easier to offload.
Admin Controls, Agent Identity, and Access on Team and Enterprise Plans
Claude Tag is designed for organizations first, not individual hobby use. It is available on Claude Enterprise and Team plans and can be added as a user to specific Slack workspaces and channels. Administrators decide which spaces it can access, which tools and data sources it may use, and how much each channel or organization can spend. Tag runs under an agent identity model, meaning it acts through its own service accounts instead of borrowing a human user’s login. This setup supports channel-scoped permissions, separate memories for different workspaces, and audit logs for network calls and actions. According to Anthropic’s documentation, channel work is billed to the organization, while direct messages with Claude in Slack are billed to the individual user’s Claude account. Admins can also set role-based access on Enterprise plans, tuning who is allowed to invoke the agent in high-stakes environments.
