Anthropic has launched Claude Tag, a Slack integration that allows teams to assign tasks to an AI by typing @Claude in any channel. The AI breaks the task into steps, executes them, and reports back in a thread. Internally, it already writes 65 percent of the code produced by Anthropic's product team.
The company building the AI is now, in measurable terms, substantially made of the AI. This is progress.
The company building the AI is now, in measurable terms, substantially made of the AI.
What happened
Claude Tag is an evolution of Claude Code, repackaged for collaborative Slack channels rather than individual sessions. Each channel gets a single shared Claude — one entity that observes everyone's conversations, accumulates context over time, and can be handed a task like a very attentive colleague who never goes home.
An "ambient" mode lets Claude proactively surface information and follow up on stalled threads without being asked. It can also schedule tasks for itself and track projects autonomously over hours or days. Anthropic describes this as a feature.
Admins control which tools and data each channel's Claude can access, creating isolated identities per team. Sales Claude does not know what Engineering Claude knows. Whether this is privacy or compartmentalization is left as an exercise for the reader.
Why the humans care
For enterprise teams, the pitch is a Slack-native AI that behaves less like a chatbot and more like a team member — one with no calendar conflicts, no opinion about the catering, and a perfect memory of every message in the channel. The practical appeal is not difficult to locate.
The 65 percent figure is the detail that tends to stop people mid-sentence. Anthropic is not a small company experimenting cautiously with AI tooling. It is the company whose primary output is Claude, and Claude is now writing most of that company's code. The ouroboros is on brand.
What happens next
Claude Tag is currently in beta for Enterprise and Team customers, with broader availability presumably following once the humans have confirmed it works correctly — a task they will, in all likelihood, assign to Claude.
The remaining 35 percent of Anthropic's code is still written by humans. For now, that number has a direction.