Playwrightess

The Playwrightess MCP server gives your AI agents a persistent Playwright environment.

Forget restarting browser sessions between commands. It maintains state through a single JavaScript interface called playwright_eval. You write standard Playwright scripts, but the context sticks around.

Need to fill forms across multiple pages? Handle authentication flows? Scrape dynamic sites without reloading? This handles it. It’s built for agents that need continuity, not one-off interactions.

Features

• ✨ Persistent browser context between API calls
• 🤖 Single playwright_eval entry point for all Playwright operations
• ⚡ JavaScript-first interface (no JSON parsing headaches)

Use Cases

• Auth workflows that don’t break: Agents log into apps once, then navigate protected routes across calls. No cookie-juggling or session resets. Regular Playwright MCP restarts the browser each time; this keeps it warm.
• Multi-step form filling: Booking flights, filing reports, tasks requiring 5+ page hops. State persistence means your agent doesn’t lose progress when switching contexts.
• Dynamic site scraping: Sites that load data via XHR after login? Maintain the session, wait for elements, and extract data across requests without re-authing.
• Debugging agent hallucinations: When your agent claims “the button exists but won’t click,” inspect the actual DOM state from the previous step. No more guessing.

How To Use It

1. Clone the MCP server and install it:

npm install
npm run build

2. Configure your MCP client (e.g., in mcp.json):

{  
  "mcpServers": {  
    "playwriter-mcp": {  
      "type": "stdio",  
      "command": "node",  
      "args": ["/your/path/to/dist/index.js"],  
      "env": {}  
    }  
  }  
}  

3. Now call playwright_eval with raw Playwright code. Example:

// First call: Log in and store session  
await playwright_eval(`  
  const { chromium } = require('playwright');  
  const browser = await chromium.launch();  
  const page = await browser.newPage();  
  await page.goto('https://app.example/login');  
  await page.fill('#email', '[email protected]');  
  await page.fill('#password', 'secret');  
  await page.click('#submit');  
  // Cookies persist for next call  
`);  

// Second call: Use the same session  
await playwright_eval(`  
  // No login needed—cookies still valid  
  await page.goto('https://app.example/dashboard');  
  console.log(await page.textContent('.welcome'));  
`); 

FAQs

Q: Why not just use regular Playwright MCP?
A: Standard Playwright MCP restarts the browser for every request. Playwrightess keeps the session alive, critical for workflows needing continuity. If your agent handles multi-page tasks, this cuts 70% of your error handling.

Q: Memory leaks? How long can sessions last?
A: Test it yourself. I ran sessions for 2 hours scraping e-commerce sites. But it’s experimental. Don’t trust it for days-long runs. Monitor memory; restart the MCP server if usage spikes.

Q: Can I use this with non-JS agents like Python?
A: Only if your agent speaks JavaScript. The playwright_eval interface demands valid JS strings. Wrap calls in your agent’s tool executor, but the payload must be JS.