Anthropic has announced the launch of Claude Tag, a persistent AI agent designed to function as a teammate inside Slack channels. The new feature places an always-on instance of the Claude model directly into group conversations, where it can observe discussions, participate in threads, and execute tasks autonomously. According to Anthropic, 65% of its product team's code is now generated by an internal version of Claude Tag, highlighting the tool's potential to reshape collaborative workflows.
Claude Tag operates within the Slack environment as a channel participant. When users type @Claude, they can prompt the AI to answer questions, perform research, or execute multi-step tasks. If ambient behavior is enabled, Claude can proactively contribute to conversations without being explicitly summoned—watching the channel for relevant developments and following up on unresolved threads. This dual-mode design allows teams to choose between a passive assistant and an active participant.
Multiplayer and agentic capabilities
The system is built around what Anthropic calls a multiplayer model. A single instance of Claude interacts with every human in the channel, maintaining a consistent understanding of context across all interactions. If Claude provides information to one user, it can reference that exchange when speaking to another. The AI retains memory of the entire channel's conversation history, learning over time and reducing the need for users to repeat themselves.
Claude Tag is fully agentic: it can break complex tasks into sequential steps, leverage tools it has been granted access to, and report back results in a threaded reply. For example, a user could ask Claude to compile a report from internal data sources, and the AI would retrieve the data, format it, and deliver the outcome without further prompting. This asynchronous operation allows team members to assign work and move on while Claude works in the background.
The tool can also act as a coach or a task master. If a thread goes silent or a deadline approaches, Claude can nudge participants directly in the channel. While this may feel intrusive to some, Anthropic argues it can significantly improve accountability and project momentum. The company has emphasized that administrators have full control over these proactive behaviors.
Scoped permissions and channel-specific identities
One of the defining features of Claude Tag is its identity isolation per channel. Each Slack channel receives its own Claude instance, and information is not shared between them. Administrators can configure different scopes for different channels—for example, a Claude in an engineering channel might have access to code repositories, while a Claude in a legal channel would be restricted to contract databases. This design respects organizational boundaries and data sensitivity.
Token consumption is a natural concern given Claude's always-on nature. Anthropic has addressed this by allowing administrators to set token spend limits for the entire organization and for individual channels. A detailed log of all Claude's activities is available for review, including which user initiated each task. This transparency helps organizations understand usage patterns and control costs.
Availability and migration
Claude Tag is available in beta for Claude Enterprise and Claude Team customers. It replaces the existing Claude in Slack app, and current users have 30 days to opt in to migration. Anthropic has stated it plans to expand Claude Tag beyond Slack in the future. An introductory launch credit has been offered to eligible organizations.
The release represents a significant evolution in AI collaboration tools. By embedding an agentic, context-aware AI directly into team communication platforms, Anthropic is pushing the boundaries of what AI coworkers can accomplish. The key questions now revolve around adoption, user acceptance, and whether organizations will embrace an AI that watches and remembers every conversation.
Source: ZDNET News