ScreenwalkMap the UI you actually built.

See every discovered screen and path. Review the real product as a whole, then copy precise changes back to your coding agent.

Try the live map

Map your first app

Actual UI, not route boxes

Inspect captured desktop and mobile screens on the same canvas as the paths between them.

Browser proof, not agent claims

See which screens rendered, which exist only in code, and which still need a recorded path.

Built for the review loop

Click a screen, say what should change and what done means, then copy a bounded handoff into your coding agent.

Screenwalk showing its synthetic fixture screens connected into product paths with screen-specific feedback

The included synthetic fixture, mapped from localhost: actual screens, entry-reachable paths, an unproven route, and a change brief in one workspace.

Your product has outgrown your mental model

AI can add a route faster than you can remember it exists. A redesign can leave an old screen reachable. An auth wall can hide the experience you meant to review. A green build—or an agent saying “it works”—does not prove the UI rendered. A route list tells you files exist; it does not show the product people can actually move through.

Screenwalk turns repository evidence and a running local app into one reviewable map.

InventorySee every discovered route and visual state, including code-only and unconnected screens.

ExperienceFollow entry-reachable branches and play captured paths without tab-spelunking.

ActionCopy findings, routes, source files, and capture context back to your coding agent.

The shortest path to a useful map

Keep your app running, then point Screenwalk at its repository and local URL:

bash

npx screenwalk /absolute/path/to/app --url http://127.0.0.1:3000

Screenwalk scans, captures, starts its local Studio, and opens the map in your browser. It does not upload your application or replace its dev server.

Run the five-minute quickstart →

Public beta

Screenwalk is available from npm as a public beta. The command runs locally and writes its evidence into the current project's .screenwalk/ directory. See beta expectations before sharing sensitive captures.