ScreenBook documentation

Start here. Each guide is short, self-contained, and ordered by how you'll actually meet the tool.

Guides — using ScreenBook

Guide	Read it when
Getting started	First contact — prerequisites, install, your first project in five minutes
The studio	A tour of every control in the shell: sidebar, the feature map, device picker, flows, review, Theme Lab, self-test
Authoring	You're building screens by hand — the full contract for screens, data, scenarios, themes, flows, strings
App mode	Embedding an EXISTING single-page app instead of (or alongside) file-per-screen features
Troubleshooting	Something's red, blank, or missing

Reference — looking things up

Reference	Covers
CLI	Every command and flag: `init` · `serve` · `manifest` · `validate` · `screenshot` · `app add` · `mcp`
AI agents & MCP	Hooking ScreenBook to Claude Code (or any MCP client) so agents author screens for you

Contributing — hacking on ScreenBook itself

Document	Covers
Architecture	How the engine works: the shell, the bridge protocol, the core, CLI, MCP server, and the self-test
Testing map	Every test layer, what it covers, how to run it, and where the gaps are

One more reference lives at the repo root: AGENTS.md — the authoring contract written for AI agents (scaffolded into every project by init; humans are welcome to read it too).