name: browser-max-automation
description: Browser automation using Playwright MCP for web testing, UI verification, and form automation. Use when navigating websites, clicking elements, filling forms, taking screenshots, or testing web applications. Supports iframe operations and complex JavaScript execution.
license: Complete terms in LICENSE.txt
metadata:
author: yamapan (https://github.com/aktsmm)

Browser Max Automation

Browser automation via Playwright MCP.

When to use

• Automating browser-based workflows or QA checks
• Verifying UI states, DOM changes, or visual regressions
• Filling forms, clicking elements, or capturing screenshots

Prerequisites

Configure .vscode/mcp.json:

{
  "servers": {
    "playwright": {
      "command": "npx",
      "args": ["@anthropic-ai/mcp-playwright@latest", "--browser", "chrome"]
    }
  }
}

Quick Reference

Basic Workflow

1. browser_navigate(url)
2. browser_snapshot → get ref
3. browser_click/type(ref)
4. browser_snapshot → verify

Commands

Navigate

mcp_playwright_browser_navigate
  url: "https://example.com"

Snapshot

mcp_playwright_browser_snapshot

Returns accessibility tree with ref values for each element.

Click

mcp_playwright_browser_click
  element: "Submit button"
  ref: "f2e123"

Type

mcp_playwright_browser_type
  element: "Username"
  ref: "f2e456"
  text: "[email protected]"
  submit: true  # Press Enter

Screenshot

mcp_playwright_browser_take_screenshot
  filename: "result.png"

Wait

mcp_playwright_browser_wait_for
  text: "Loading complete"  # or
  time: 3

Tabs

mcp_playwright_browser_tabs
  action: "list" | "new" | "close" | "select"
  index: 0

Advanced

iframe Operations

async (page) => {
  const frame1 = page.locator('iframe[name="Content"]').contentFrame();
  const frame2 = frame1.locator('iframe[title="Player"]').contentFrame();
  await frame2.getByRole("radio", { name: "Option A" }).click({ force: true });
  return "Selected";
};

force: true

Use when element is covered by another (e.g., SVG overlay):

await element.click({ force: true });

When browser_run_code is disabled

Use snapshot + click instead:

browser_snapshot → get ref → browser_click(ref)

Reference