Claude Tag Beta Release

Claude Tag transforms AI from a private chat into a collaborative teammate

Claude Tag is a new beta feature for Claude Enterprise and Team customers that integrates Claude directly into Slack as a team member. Unlike traditional AI chatbots that operate in isolated 1:1 sessions, Claude Tag is a "multiplayer" agent that exists within shared channels, allowing multiple team members to delegate tasks to a single AI instance, track its progress in real-time, and build a shared body of context over time.

Core Capabilities of Claude Tag

Claude Tag shifts the AI interaction model from simple prompting to asynchronous task delegation.

Multiplayer Collaboration

Within a Slack channel, there is one shared instance of Claude. This allows any team member to see what the agent is currently working on and pick up a conversation where a previous colleague left off, mirroring the collaborative nature of a human teammate.

Persistent Context and Learning

Claude Tag builds context by remembering relevant information from the channels it is permitted to join. This eliminates the need for users to repeatedly provide the same background information. It can also learn from other authorized Slack channels and external data sources to acquire the tacit knowledge necessary for complex organizational tasks.

Proactive "Ambient" Behavior

When enabled, Claude Tag can take initiative by proactively updating the team on relevant information from its connected tools and channels. It can also follow up on unresolved threads or tasks that have gone quiet without being completed.

Asynchronous Execution

Users can assign Claude Tag a task and leave the platform while the agent works. Claude can plan and execute multi-stage projects autonomously over hours or days, responding in a Slack thread once the work is finished.

Implementation and Governance

Anthropic has designed Claude Tag with strict administrative controls to manage data access and costs.

• Scoped Identities: System administrators define which tools and data sources Claude can access on a per-channel basis. This creates separate "identities" for different functions; for example, a Claude instance scoped for sales will not share memories or data with an instance scoped for engineering.
• Cost Management: Administrators can set token spend limits for the entire organization or for specific individual channels.
• Auditability: A full log of all actions performed by @Claude, including who requested each task, is available for review.
• Technical Requirements: Claude Tag currently works with Opus 4.8.

Internal Impact at Anthropic

Anthropic reports that Claude Tag is already deeply integrated into its own internal workflows. According to the company, 65% of its product team's code is created by an internal version of Claude Tag. Beyond engineering, the tool is used to analyze product metrics, process support tickets, and identify the root causes of software bugs.

Community Perspectives and Technical Concerns

While the announcement highlights productivity gains, the developer community on Hacker News raised several critical concerns regarding security, cost, and reliability.

Security and Permissions

Critics argue that mapping AI permissions to Slack channel memberships is inherently risky. One user noted that permissions rarely align perfectly with channel members, suggesting that agents must either be "dumbed-down" to the lowest common denominator of access or treated as human employees with full liability.

Data Privacy

Some users expressed significant concern over granting a third-party company a "direct pipe" into sensitive internal discussions, viewing the integration as a potential security risk for enterprise data.

Cost and Token Consumption

There is concern that Claude Tag will be a "token guzzler," as it may need to parse large volumes of messages across multiple channels to maintain context. Users also pointed out that Claude Tag uses token-based pricing that is separate from existing Claude subscriptions.

Memory Reliability

Some users reported issues with AI "learning" from incorrect or experimental data, leading the model to make false assumptions that persist in its memory until manually cleared.

"The most important difference from other products: @Claude is multiplayer. Within a given Slack channel, there’s one Claude that interacts with everyone... This makes tagging Claude very different from working within a single chat... it’s much more like interacting collaboratively with a teammate."

"I read this as 'Claude integration to Slack is now billed as API usage'."