Anthropic launches Claude Tag: Collaborative AI integrated into Slack | Keryc
Claude Tag arrives as a different way to work with artificial intelligence inside your teams: think of a member of a Slack channel you can tag to delegate tasks, follow processes, and work in the background while you do other things. Sounds like magic? No: it’s a mix of contextual memory, administrative permissions, and automation built for real teams.
What is Claude Tag
Claude Tag is the new way to use @Claude as a collaborator inside Slack. You invite it to specific channels, give it access to particular tools and data, and anyone in that channel can tag it to request tasks. Instead of a private chat per task, it acts like a teammate that remembers channel context and can plan and execute steps over time.
At Anthropic they already use this approach: 65% of the product team’s code is generated with their internal version of Claude Tag.
Key advantages
Multiplayer: there’s a single instance of @Claude per channel, accessible to all members. That enables continuity across different people.
Learns over time: if you grant permission, @Claude can learn from the channel history and other sources to avoid repeated explanations.
Takes initiative: with the “ambient” mode on, it will notify you about relevant info or follow up on its own.
Asynchronous work: you request a task and @Claude follows it over hours or days without you having to babysit it.
How it works in practice
Asking @Claude for something is simple: tag it and give the instruction in natural language. It then breaks that request into stages, uses the authorized tools, and posts the result in a Slack thread. You can also send direct messages for private queries, where it will use only your personal connectors.
Admins control everything: which channels @Claude can see, which tools and data it can use, and even spending limits per token. Each instance is scoped to the defined channel or set of channels, which prevents memories and accesses from mixing between teams (for example, sales separated from engineering).
Concrete use cases
Product managers delegate metric searches or data analysis and then review the results in the thread.
Support teams ask @Claude to prioritize and summarize pending tickets so they can make faster decisions.
Engineering uses @Claude to investigate bugs, generate code, or run small diagnostic pipelines.
Doesn’t that sound useful for founders who need to optimize repetitive tasks without hiring more people? Exactly.
Security, control, and costs
Anthropic designed Claude Tag with corporate environments in mind: fine-grained permissions, activity logs, and spending limits per organization and per channel.
Admins can review a task log showing who requested each action.
@Claude’s memories are isolated by channel profiles, preventing leaks between areas.
Migration from the previous Claude app in Slack is optional: admins have 30 days to opt into the change.
Availability and requirements
Claude Tag launches today in beta for Claude Enterprise and Team customers and runs on Opus 4.8. Activation requires pairing Claude Tag with your Slack workspace, granting access to the necessary tools, setting a monthly spending cap, and testing it in a private channel.
Anthropic offers initial credit for eligible organizations that want to try the service before committing larger budgets.
Final reflection
This isn’t just another automation app: it’s a bet on integrating AI as a team member, not an isolated assistant. What would change in your day-to-day if you could delegate searches, follow-ups, and even parts of code to a teammate that works in parallel and learns from context? Maybe the way we organize collective work will change faster than we think.
