Anthropic brings Claude Tag, its AI, to Slack

Anthropic has launched Claude Tag, a version of Claude that can integrate into Slack as another member of the team. Instead of opening a separate chat, you can mention @Claude in a channel, assign it a task, and let it complete the work using the tools, data, and codebases the company authorizes.

The feature is available in beta starting today for Claude Enterprise and Team customers. For now, it works in Slack and replaces Claude’s previous app for the platform.

A teammate inside Slack

Claude Tag can receive an instruction written in plain language, break it into several steps, and work with connected resources. When it finishes, it replies in a thread with what it has done or created.

For example, a team can ask it to review product metrics, analyze support tickets, investigate a software bug, or prepare changes to a codebase. The idea is to keep the work in the same place where the team already communicates, instead of jumping between chats, documents, and different tools.

The feature is shared, too. Each channel has a Claude that multiple people can interact with, and everyone can see its work and continue the conversation from where another teammate left off. Anthropic describes this as a way to collaborate that is closer to working with a team member than using an individual chatbot.

Claude remembers, acts, and works in the background

Claude Tag can retain relevant information from the channels it can access, so users do not have to repeat the context in every request. It can also learn from other channels and data sources if it receives permission to do so. Anthropic clarifies that it does not report on private channels it cannot access.

The tool also includes three capabilities that set it apart from a conventional chat:

• It can surface relevant information and resume unresolved threads or tasks when “ambient” behavior is enabled.
• It can work asynchronously while the team focuses on other priorities.
• It can schedule tasks to continue a project over hours or days.

You can also write to it by direct message. In that case, Claude responds privately using the personal tools and connectors the user has configured.

Anthropic says this model is already being used internally. According to the company, 65% of the product team’s code is currently created with its internal version of Claude Tag. Its use has also expanded to metrics, support, and error analysis.

The company controls the permissions

Claude Tag is designed for organizations that need precise control over what the AI can access and do. Administrators decide which tools and data are available to Claude and in which channels it can use them.

This makes it possible to create separate configurations for different teams. A Claude assigned to sales does not share its memories with another configured for engineering, nor can it give engineers access to data or tools from the sales department.

Administrators can also:

• Set spending limits for the entire organization and for each channel.
• Review a record of the actions Claude has taken.
• See who requested each task.
• Test the tool in a private channel first.

To get started, companies must connect Claude Tag to their Slack workspace, grant it access to the necessary tools, set a monthly spending limit, and check that it works as expected. The feature runs on Opus 4.8.

For you, the main change is that asking AI for help is no longer an individual activity. A team can assign tasks from a shared conversation, track progress, and reuse the accumulated context. The challenge will be deciding which permissions to grant and reviewing the cost and actions it can execute. Anthropic plans to bring Claude Tag to other workplaces, but has not yet specified when or on which platforms.