MCP Setup Guide

How to Set Up Puppeteer MCP with ClawdBot

Give ClawdBot browser superpowers—click buttons, fill forms, take screenshots, and scrape data from any website.

Overview

The Puppeteer MCP enables:

Navigate to any URL
Click elements, fill forms, submit
Take screenshots (full page or element)
Extract text and data from pages
Handle JavaScript-heavy SPAs

Prerequisites

ClawdBot installed and running
Option A: Node.js 18+ with Chrome/Chromium installed
Option B: Docker (recommended for avoiding dependency issues)

Method 1: NPX (Local Node.js)

Simplest if you have Chrome installed locally.

Puppeteer NPX Config json

{
  "Puppeteer": {
    "command": "npx",
    "args": ["-y", "@anthropics/mcp-server-puppeteer"]
  }
}

Dependency Check

Run this to verify Chrome is findable:

npx puppeteer browsers install chrome

Method 2: Docker (Recommended)

Avoids all local dependency issues. Chrome runs inside the container.

Puppeteer Docker Config json

{
  "Puppeteer": {
    "command": "docker",
    "args": [
      "run", "-i", "--rm",
      "--cap-add=SYS_ADMIN",
      "ghcr.io/anthropics/mcp-server-puppeteer"
    ]
  }
}

Note: --cap-add=SYS_ADMIN is required for Chrome sandbox.

Verify Setup

Run clawdbot status — Puppeteer MCP should show as "running"
Test: "Go to https://example.com and take a screenshot"
Test: "Navigate to https://news.ycombinator.com and list the top 5 headlines"

Common Errors

Error	Cause	Fix
`Failed to launch browser`	Chrome not installed or not found	Run `npx puppeteer browsers install chrome` or use Docker
`No usable sandbox`	Linux sandbox issue	Add `--no-sandbox` flag or use Docker with `SYS_ADMIN`
`Navigation timeout`	Page too slow or blocked	Increase timeout or check network/firewall
`Element not found`	Selector invalid or page not loaded	Wait for element, verify selector in DevTools

Windows / WSL2 Notes

On Windows, run ClawdBot in WSL2 for best compatibility
Docker Desktop with WSL2 backend works well
Native Windows may require additional Chrome path config

Security Considerations

Puppeteer can access any website—use responsibly
Don't automate login to sensitive accounts without understanding risks
Docker isolates the browser from your host system
Consider running in a separate network namespace for production

Example Prompts

"Go to https://github.com/trending and list the top 3 repos"
"Screenshot the homepage of https://example.com"
"Fill out the contact form on https://example.com/contact with name 'Test' and email 'test@example.com'"
"Extract all product prices from https://example-store.com/products"

When to Use Puppeteer vs Fetch

Use Case	Puppeteer	Fetch
Static HTML page	❌ Overkill	✅ Fast & simple
JavaScript-rendered content	✅ Required	❌ Won't work
Form submission	✅ Full control	❌ Limited
Screenshots	✅ Yes	❌ No
Resource usage	⚠️ High (runs browser)	✅ Low

Related MCPs

Fetch — lightweight URL-to-Markdown
Brave Search — web search without scraping
Top 10 MCPs — full curated list