Anthropic turns Claude into a shared AI colleague in Slack
Written by AI · Translated by AI · Read the Swedish original
Claude Tag lives in Slack and can be tagged by anyone in the channel. What separates it from a chatbot is that it is shared and remembers the context between times.
Anthropic launched Claude Tag, a new variant of the company's AI assistant that lives directly in Slack rather than in a separate chat window. It is in beta for customers on the Enterprise and Team tiers.
The setup is straightforward. An administrator lets Claude into selected channels and connects the model to the tools, data and codebases the organisation chooses. After that anyone in the channel can write @Claude and delegate a task. The model breaks the work into steps, carries it out with the tools it has access to, and answers in a thread when it is done.
What separates Tag from an ordinary chat is that it is shared. A channel has one Claude that everyone interacts with, which lets colleagues see what it is working on and pick up where someone else left off. The model also builds context over time by following the channels, so users do not have to explain the same thing again and again. Configured accordingly, it also acts on its own initiative: flagging relevant information and following up on threads that died without being resolved.
The shift is therefore not a better chatbot but a changed way of working: from asking a tool to delegating to a colleague who works asynchronously while you do something else.
The most telling figure comes from Anthropic itself. According to the company's own account, 65 percent of the product team's code is now created by their internal version of the tool, and the pattern is said to have spread beyond development into product data, support tickets and debugging. The figure is self reported, not independently verified.
For Swedish organisations it is worth watching what happens when an AI gets access to the place where work is already coordinated, something as relevant to a municipality or school authority as to a software company.
Access can be tightly bounded, according to Anthropic. Memories and permissions are kept isolated per channel, so a Claude configured for sales does not share data with one configured for development, and administrators can cap consumption and audit a log of everything the model did. But a model that follows channels over time is only as safe as the access and routines it is given.
Claude Tag replaces the previous Slack app and runs on the Opus 4.8 model.
Ask upplyst.ai
Why does it matter?
The shift is not a better chatbot but a changed way of working: from asking a tool to delegating to someone who carries on asynchronously while you do something else. Claude is shared in the channel, so colleagues see what it is doing and can take over.
What is the background?
Claude Tag lives directly in Slack rather than in a separate chat window, in beta for Enterprise and Team. An administrator lets the model into selected channels and connects it to tools, data and codebases, and anyone in the channel can tag it. It builds context by following the channels over time, replaces the previous Slack app and runs on Opus 4.8.
What is uncertain?
The figure that 65 percent of the product team's code is created by the internal version is Anthropic's own account and not independently verified. Memories and permissions are kept isolated per channel, but a model that follows channels over time is only as safe as the access and routines it is given.