Anthropic has introduced Claude Tag, a new way for teams to work with its AI model inside Slack. The company is launching the feature first in beta for Claude Enterprise and Team customers, positioning Slack as a central place where groups can delegate work to Claude alongside human teammates.
With Claude Tag, administrators can grant Claude access to selected Slack channels and connect it to approved tools, data sources and codebases. Once configured, users in those channels can tag @Claude to assign tasks, ask for help or request follow-up work. Anthropic said the system is designed so Claude can break requests into steps, carry out the work using its connected tools, and report back in the Slack thread when it is finished.
The company says the product is meant to extend the capabilities of Claude Code and make the model more useful in shared work settings. Anthropic described tagging Claude as one of its own main workflows internally. According to the company, 65% of the product team’s code is now created using its internal version of Claude Tag. It also said teams are using the tool beyond software development for tasks such as tracking product metrics, working through support tickets and investigating bugs.
Anthropic is pitching Claude Tag as more than a simple chat interface. In a Slack channel, the company says there is one Claude that can interact with everyone in the conversation, allowing different team members to see its work and continue from the same context. That shared setup is intended to make the experience feel closer to a team member than a private AI chat.
The system also remembers relevant details from the channels it is allowed to access, which Anthropic says reduces the need for users to repeat background information. If given permission, Claude can also learn from other channels and data sources, while remaining restricted from private channels it cannot access.
Another feature Anthropic is highlighting is what it calls ambient behavior. When enabled, Claude can proactively surface information it thinks a team may need, point out relevant updates from the channels and tools it is connected to, and follow up on unresolved threads or dormant tasks.
The company also says Claude can work asynchronously, meaning users can hand off a task and return later while the model continues working. In some cases, Claude can even schedule work for itself and pursue tasks over the course of hours or days.
Anthropic said Claude Tag was built with organizational controls in mind. System administrators decide which tools and information Claude can access and where it can operate, with memory and permissions scoped to specific uses. The company gave an example of separate Claude setups for sales and engineering, each limited to its own data and tools.
Administrators can also set spending limits for the organization and for individual channels. Anthropic said they will be able to review logs of what Claude did and who requested each task.
Claude Tag replaces the existing Claude in Slack app. Anthropic said administrators can opt in to migration within 30 days. Eligible Enterprise and Team customers will also receive an introductory launch credit so their organizations can test the product.
The feature works with Opus 4.8, and Anthropic says it plans to expand availability beyond the current beta over time. For now, the company is presenting Claude Tag as a first step toward bringing Claude into more of the places where teams collaborate.