electron-playwright-mcp-docs

A documentation-only reference for building a multi-window Electron application, driving it end-to-end with a Playwright-based MCP (Model Context Protocol) server, and letting an AI agent author and iterate the resulting tests through that server.

No source code. No package.json. No build. This repository contains only Markdown, Mermaid diagrams, and example CI workflow files.

Licensed under CC-BY-4.0.

Suggested reading order:

flowchart TD
    overview["<b>00 — Overview</b><br/>problem statement and scope"]
    arch["<b>01 — Electron architecture</b><br/>processes, IPC, two-window model"]
    mcp["<b>02 — MCP server</b><br/>tool vocabulary, injection patterns"]
    agent["<b>03 — AI-agent workflow</b><br/>authoring tests through the MCP"]
    ci["<b>04 — CI examples</b><br/>GitHub Actions and GitLab workflows"]
    overview --> arch --> mcp --> agent --> ci

Audience

• Electron developers writing real end-to-end tests for their desktop app.
• Senior engineers evaluating a multi-window Electron + Playwright + MCP pattern before investing in it themselves.

Table of contents

Section	Status	Contents
00-overview/	Phase 1	Problem statement, scope, non-goals, reading order
01-electron-architecture/	Phase 1	Process topology, IPC surface, two-window model, build stack; ADR-001 (companion window pattern) and ADR-002 (build stack)
02-mcp-server/	Phase 2	MCP server architecture, tool catalogue (18 tools), injection patterns (6), ADR-003 (stdio transport)
03-ai-agent-workflow/	Phase 2	Agent loop, worked cross-window test example, fixture-based testing vs POM, ADR-004 (fixtures over POM)
04-ci-examples/	Phase 2	Example GitHub Actions and GitLab CI workflows (not extracted from source)

Conventions

• Diagrams are Mermaid, colocated with the prose that introduces them.