Claude Code adds /clone-website with Chrome MCP and git worktrees
An open-source Claude Code template now clones websites from a single /clone-website command using Chrome MCP, design-token capture, and parallel git worktrees. It packages front-end recreation into a repeatable flow, but current proof comes from repo demos rather than broad field use.
TL;DR
- Linus Ekenstam surfaced an open-source Claude Code template that turns a single
/clone-websitecommand into a full site recreation flow, with the public demo framed around "pixel-perfect" cloning demo thread. - The thread says the skill inspects a target site through Chrome MCP, extracts design tokens and assets, writes component specs, and then rebuilds sections in parallel rather than step by step workflow post.
- The implementation detail that makes it feel production-minded is isolated git worktrees that auto-merge after each builder agent finishes its section workflow post.
- Ekenstam also said he was about to speak with the builder, Ben, which suggests more details may follow; for now, the evidence is still mostly the repo preview and demo thread rather than broad creator adoption follow-up post.
How does the cloning workflow actually work?
The pitch is not "AI can mock up a landing page." It is a reusable cloning pipeline. In Ekenstam's description, /clone-website first uses Chrome MCP to inspect the live site, then captures design tokens and assets, generates component specs, and hands sections off to parallel builder agents. The README preview in the post labels it an "AI Website Clone Template" for reverse-engineering sites into pixel-perfect Claude Code rebuilds README preview README screenshot.
What is still missing is field evidence. The public posts show the workflow architecture and an interview setup with Ben, but not a slate of finished creator projects, failure cases, or limits across different site types. A supporting repost praising the idea adds enthusiasm, not new technical detail reaction repost.