BrowserAuto

AI-powered E2E testing agent — point it at any web app, and it automatically analyzes the UI, generates a test plan, and executes tests in a real browser.

Features

• Automatic app analysis — AI explores your web app's source code to understand routes, forms, user flows, and API endpoints
• Test plan generation — produces structured, editable test suites with prioritized cases
• Real browser execution — runs tests in Chromium via Playwright with screenshots, console logs, and network error capture
• MCP server — expose the run_tests tool over HTTP so Claude Code or any MCP client can trigger tests programmatically
• Desktop app — native Electron UI with a dark, macOS-inspired interface

Prerequisites

• Node.js 18+
• Bun (package manager & script runner)

Getting Started

# Install dependencies
bun install

# Start in development mode
bun run dev

# Build for production
bun run build

# Build distributable for macOS
bun run build:mac

How It Works

1. Select a project — pick a local project folder or enter a live app URL
2. Analyze — the AI agent explores your source code using filesystem tools and produces a structured analysis (routes, forms, user flows, API endpoints)
3. Generate test plan — based on the analysis, the agent creates prioritized test suites with concrete steps (navigate, click, fill, assert, screenshot, etc.)
4. Execute — tests run in a real Chromium browser; each step is executed, screenshots are captured, and console/network errors are collected
5. Review results — see pass/fail status, screenshots, logs, and errors for every test case

MCP Server / Claude Code Integration

BrowserAuto exposes an MCP server on http://127.0.0.1:3100/mcp that any MCP-compatible client can use.

Claude Code config

Add this to your Claude Code MCP settings:

{
  "mcpServers": {
    "browserauto": {
      "type": "url",
      "url": "http://127.0.0.1:3100/mcp"
    }
  }
}

The server exposes a run_tests tool that accepts an app URL and an array of test cases, then returns results with screenshots, console logs, and error details.

Tech Stack

Layer	Technology
Desktop framework	Electron
Frontend	Svelte 5 (runes)
Browser automation	Playwright (Chromium)
AI integration	Vercel AI SDK + OpenRouter
MCP transport	HTTP via @modelcontextprotocol/sdk
Schema validation	Zod
Build tooling	electron-vite, Vite, TypeScript
Code quality	ESLint, Prettier

Contributing

See CONTRIBUTING.md for development setup and contribution guidelines.

License

MIT — Copyright (c) 2026 Daniel Ionut