Domscribe

About Domscribe

Domscribe closes a specific and very common failure in agent-assisted frontend work: the agent can read every source file in the repository but cannot see what any of it renders as, so when you ask it to change the element you are looking at, it has to guess which file owns that element. The project's framing is that every agent-assisted UI change is a search problem before it is a coding problem — ask for a blue hero button, and the agent greps for `button`, finds twelve files, and edits the wrong one. Domscribe fixes this by running at build time: it walks JSX and Vue templates, assigns every element a stable ID, and writes a manifest mapping each ID to its exact file, line and column. The agent reads the manifest and resolves the element on the first try. The bridge runs in both directions. A developer can click a DOM element in the browser to tell the agent what to change, and the agent can query any source location to see what it currently renders as in the live DOM before touching it. It is framework-agnostic across React, Vue, Next.js, Nuxt, Vite, Webpack and Turbopack, works with Claude Code, Cursor, GitHub Copilot, Gemini CLI, Amazon Kiro or any MCP-compatible agent, and has zero production impact.

Key Features

Tags