← Back to news

Claude Tag

anthropic.com|57 points|22 comments|by adocomplete|Jun 23, 2026

Introducing Claude Tag

Date: June 23, 2026
Source: Anthropic

Anthropic is excited to unveil Claude Tag, a transformative approach to team-based AI collaboration. Initially launching on Slack, Claude Tag allows the AI to integrate directly into your workspace as a functional team member.

By granting Claude access to specific channels and linking it to your preferred data sources, tools, and codebases, your team can simply tag @Claude to delegate complex assignments, freeing up human employees to focus on high-level strategy.

The Evolution of AI Collaboration

Claude Tag represents a significant leap forward from Claude Code. It shifts the model from a reactive tool to a proactive teammate.

"Tagging @Claude is now one of the primary drivers of productivity at Anthropic."

The impact is already evident internally. For instance, the volume of code produced by the product team via their internal version of Claude Tag is: Code Generation Rate=65%\text{Code Generation Rate} = 65\%

Beyond just writing code, the team utilizes @Claude to:

  • Analyze and track product metrics.
  • Resolve customer support tickets.
  • Identify the root causes of complex software bugs.

Claude Tag Workflow Diagram

How Working with @Claude Differs

If you have experience with Cowork or Claude Code, the interface will feel intuitive. You provide a request in plain language, and Claude decomposes the task into manageable stages, executing them using its available toolset. Once finished, it posts the results within a Slack thread.

However, Claude Tag introduces a "multiplayer" dynamic. Unlike a private chat, there is one shared Claude instance per channel.

Comparison: Standard AI Chat vs. Claude Tag

FeatureStandard AI ChatClaude Tag
VisibilityPrivate/IndividualShared/Transparent
ContextReset per sessionPersistent per channel
CollaborationSingle-userMultiplayer (Team-wide)
ProactivityReactive onlyAmbient & Proactive

Context and "Ambient" Intelligence

Claude doesn't just process prompts; it observes. By following channel conversations, it builds a deep reservoir of tacit knowledge, meaning users don't have to repeat instructions no longer need to provide exhaustive context for every request.

When "Ambient Behavior" is toggled on, Claude becomes a proactive sentinel:

  1. It monitors connected tools and channels for relevant updates.
  2. It flags critical information for the team.
  3. It follows up on stagnant threads to ensure resolution.

Furthermore, Claude can autonomously schedule its own work, pursuing long-term projects over several days. For those needing privacy, direct messages remain available, utilizing personal connectors and tools.

Administrative Control & Security

Designed for the enterprise, Claude Tag offers granular permissioning. System administrators define exactly which tools and data the model can access within specific channels.

This creates scoped identities. For example: if (channel == "Sales") { access = sales_data; memory = sales_context; } if (channel == "Engineering") { access = codebase; memory = eng_context; }

This ensures that an engineering-scoped Claude cannot leak sales data or access restricted sales tools.

Management Features

  • Spend Tracking: Admins can set token limits for the entire organization or specific channels.
  • Audit Logs: A complete history of @Claude's actions and the users who initiated them is maintained.

Getting Started

Claude Tag is currently in beta for Claude Team and Claude Enterprise customers. To begin implementation, follow this checklist:

  • Pair Claude Tag with your Slack workspace.
  • Configure access to necessary tools and data.
  • Establish a monthly spending limit for the organization.
  • Conduct a trial run in a private channel to verify functionality.

Note: Eligible organizations will receive an introductory launch credit to facilitate company-wide testing.


Related Updates

  • Global Expansion: Anthropic has opened a new office in Seoul and formed several partnerships within the Korean AI ecosystem.
  • Regulatory Compliance:

    Notice: Per US government export control directives, all access to Fable 5 and Mythos 5 has been suspended.

  • Transparency: Review the latest findings from the first Anthropic Public Record.