Anthropic Revolutionizes Slack with Claude Tag, a Collaborative AI Agent

Anthropic Integrates Claude Tag into Slack for Optimized AI Collaboration

Anthropic continues to expand the use of its AI assistant, Claude, in professional environments. After integrating Claude into platforms such as Slack, Figma, and Canva earlier this year, the company now unveils Claude Tag. This new feature transforms Claude into a permanent teammate in Slack, replacing the previous app, Claude in Slack, which was removed on August 3. Claude Tag is presented as an evolution of Claude Code, offering a continuous and proactive presence in messaging.

A Collaborative AI Agent Accessible to the Entire Team

Claude Tag operates in a simple yet effective manner. Administrators can grant Claude access to specific Slack channels, connect it to various tools, databases, and even specific codebases. Each member of the channel can then mention @Claude to assign tasks. The AI breaks down the request into steps, uses the available tools to process these steps, and provides the results in a discussion thread. This feature currently relies on version Opus 4.8.

Anthropic highlights four distinctive features of Claude Tag compared to previous uses of the AI:

• Multiplayer: A single Claude is shared within a channel, allowing all members to see ongoing tasks and pick up a conversation where a colleague left off.

• Context Memory: Claude tracks exchanges in the channel, accumulating information about ongoing work. With permission, it can also access other channels and data sources, although it cannot extract content from private channels.

• Initiative: In "ambient" mode, Claude can intervene on its own to signal useful information or revive a discussion thread that has gone unanswered.

• Asynchronous: Claude can work in the background while the team focuses on other tasks, and schedule tasks to be executed over several hours or days.

Governance and Control of AI

Anthropic emphasizes the control that companies retain over the use of Claude Tag. Administrators can determine which tools, information, and channels are accessible to Claude. Each instance of Claude is limited to its specific scope: for example, a Claude configured for the sales department will not share its data with the engineering department. Managers can also limit token spending at the organization or specific channel level and review a detailed log of actions taken, including the identity of the requester. It is also possible to send private messages to Claude, which will then respond using the user's personal tools.

Availability of Claude Tag

Claude Tag is currently available in beta for clients of the Claude Enterprise and Team offerings. It replaces the Claude in Slack app, which was removed on August 3, and organizations have 30 days to migrate if they wish. A launch credit is offered to eligible organizations to facilitate this transition.

Towards a Permanent AI Teammate

Claude Tag succeeds Claude in Slack, which already allowed users to request Claude for occasional help or to pass coding requests to Claude Code. However, Claude in Slack could not maintain a lasting context between exchanges. With Claude Tag, this channel memory is now possible, offering a shared identity for the entire team. This change alters the way teams interact with AI, moving from isolated interactions to a continuous dialogue with a "teammate" who understands the project's context. The ability to delegate tasks asynchronously also allows for managing multiple tasks in parallel, a mode of operation that Anthropic has adopted internally. According to the company, their internal version of Claude now produces 65% of their product team's code and has contributed to the development of Claude Tag itself. For now, this feature is reserved for Enterprise and Team subscriptions.