<p>The Playwright MCP server gives AI agents direct browser control.
Primary use case: verifying rendered output during development.</p>
<h2 id="key-tools">Key Tools</h2>
<ul>
<li><code>browser_navigate</code>: load a URL</li>
<li><code>browser_take_screenshot</code>: capture the current viewport</li>
<li><code>browser_snapshot</code>: read the accessibility tree (structured text)</li>
<li><code>browser_click</code>, <code>browser_fill_form</code>: interact with elements</li>
<li><code>browser_wait_for</code>: wait for selectors or navigation</li>
<li><code>browser_console_messages</code>: check for JS errors</li>
</ul>
<h2 id="verification-workflow">Verification Workflow</h2>
<ol>
<li>Start dev server (<code>bin/start-dev-bg.sh</code>)</li>
<li>Navigate to the page under test</li>
<li>Take screenshot to visually inspect</li>
<li>Use snapshot to programmatically check text/structure</li>
<li>Fix issues, rebuild if needed, repeat</li>
<li>Close browser and kill dev server when done</li>
</ol>
<h2 id="related-blog-posts">Related Blog Posts</h2>
<ul>
<li><a href="/playwright-mcp.html">Playwright MCP: Claude Code Verifying Its Own Work</a>:
setup guide and workflow</li>
</ul>


Browser automation MCP server. Enables AI agents to navigate pages, take screenshots, inspect accessibility trees, and verify rendered output.

Kyle Pericak

"It works in my environment"

Playwright MCP

Key Tools

Verification Workflow