AI Browser Automation - MCP Server for Claude & ChatGPT Enable AI assistants to control web browsers through natural language. This Apify Actor is a Model Context Protocol (MCP) server that gives Claude, ChatGPT, and other AI agents 17 browser automation tools to scrape data, fill forms, take screenshots, and automate any web task. --- ## What is MCP Browser Automation? This Actor transforms your AI assistant into a browser automation expert. Instead of writing complex scripts, describe what you need in plain English—your AI handles everything using a real Chrome browser. Example conversation with Claude: > "Go to example.com, extract all product prices, and take a screenshot of the pricing table" Claude automatically uses the right tools (navigate → extract → screenshot) and returns the results. --- ## Key Features | Feature | Description | | ----------------------- | --------------------------------------------------------------------------------- | | 17 Browser Tools | Complete automation: navigate, click, type, screenshot, extract, scroll, and more | | Session Persistence | Browser state maintained across multiple tool calls | | Concurrent Browsers | Up to 10 parallel browser instances for faster scraping | | Real Chrome Engine | Full JavaScript support—handles React, Vue, Angular, SPAs | | Dataset Storage | Large outputs (screenshots, HTML) automatically stored in Apify Dataset | | No Coding Required | Describe tasks in natural language to your AI assistant | --- ## What Data Can You Extract? | Data Type | Examples | Use Cases | | ------------------- | ---------------------------------- | --------------------------------------- | | Product Data | Prices, descriptions, reviews | Price monitoring, market research | | Contact Info | Emails, phone numbers, addresses | Lead generation, prospecting | | Content | Articles, posts, comments | Content aggregation, research | | Screenshots | Full page or specific elements | Visual monitoring, documentation | | Dynamic Content | Lazy-loaded, infinite scroll, SPAs | Modern web apps, JavaScript-heavy sites | --- ## Quick Start: Connect to Claude Desktop ### Step 1: Start the Actor on Apify 1. Click "Try for free" on the Actor page 2. Keep default settings (or adjust timeout for slow sites) 3. Click "Start" 4. Copy the Actor endpoint URL from run details ### Step 2: Configure Claude Desktop Find your Claude config file: - macOS: ~/Library/Application Support/Claude/claude_desktop_config.json - Windows: %APPDATA%\Claude\claude_desktop_config.json - Linux: ~/.config/Claude/claude_desktop_config.json Add this configuration: json { "mcpServers": { "puppeteer-mcp": { "type": "http", "url": "https://meysamazing--puppeteer-mcp.apify.actor?token=apify_api_..." } } } or run the CLI: shell claude mcp add --transport http puppeteer-mcp https://meysamazing--puppeteer-mcp.apify.actor?token=apify_api_... Restart Claude Desktop. Done! ### Step 3: Start Automating Try these prompts with Claude: - "Navigate to reddit.com and extract the top 5 post titles" - "Go to example.com/pricing and take a screenshot" - "Fill the contact form with name 'John' and email 'john@test.com'" --- ## All 17 Browser Tools Your AI automatically selects the right tool based on your request. ### Navigation | Tool | What It Does | | --------------- | ----------------------------------- | | navigate | Go to any URL, wait for page load | | goBack | Navigate back in browser history | | goForward | Navigate forward in browser history | | getCurrentUrl | Get current page URL and title | ### Interaction | Tool | What It Does | | ---------- | --------------------------------------------------- | | click | Click elements (CSS selector, XPath, or text match) | | type | Type text into inputs with optional delay | | hover | Hover over elements to trigger tooltips/menus | | scroll | Scroll page or scroll to specific element | | select | Choose options from dropdown menus | | pressKey | Press keyboard keys (Enter, Tab, Escape, etc.) | ### Data Extraction | Tool | What It Does | | ----------------- | -------------------------------------------------------- | | screenshot | Capture full page or element screenshots | | getContent | Get page content as HTML, text, or markdown | | extract | Extract text/attributes from single or multiple elements | | evaluate | Execute custom JavaScript on the page | | waitForSelector | Wait for elements to appear before acting | ### Session Management | Tool | What It Does | | ---------------- | -------------------------------- | | listSessions | View all active browser sessions | | destroySession | Close a specific browser session | --- ## Input Configuration Customize the Actor for your needs: | Setting | Default | Description | | ------------------ | --------- | ---------------------------------------------- | | headlessMode | true | Run browser without UI (disable for debugging) | | defaultTimeout | 30000 | Max wait time per operation (ms) | | maxConcurrency | 1 | Parallel browser instances (1-10) | | logLevel | info | Logging verbosity (debug/info/warn/error) | | sessionTimeoutMs | 1800000 | Idle session timeout (30 min default) | | viewportWidth | 1280 | Browser viewport width (px) | | viewportHeight | 720 | Browser viewport height (px) | Recommended for slow websites: json { "defaultTimeout": 60000, "maxConcurrency": 1 } Recommended for faster scraping: json { "defaultTimeout": 30000, "maxConcurrency": 5 } --- ## Output Examples ### Navigation Result json { "sessionId": "abc-123", "url": "https://example.com", "title": "Example Domain", "status": 200 } ### Data Extraction Result json { "sessionId": "abc-123", "selector": ".product-price", "count": 5, "values": ["$19.99", "$24.99", "$29.99", "$34.99", "$39.99"] } ### Screenshot Result (Large Images Stored in Dataset) json { "sessionId": "abc-123", "format": "png", "sizeKB": 450, "storedInDataset": true, "datasetItemUrl": "https://api.apify.com/v2/datasets/.../items" } --- ## How Much Does It Cost? This Actor runs on Apify's platform. Pricing will be Pay Per Event (PPE) based on tool usage. Current model (until PPE is finalized): - Compute Units: ~$0.50/hour of browser runtime - Free tier: 5 hours/month included with Apify account Estimated costs: | Usage | Operations | Estimated Cost | | ------ | ---------------- | -------------- | | Light | ~50 pages/day | Free tier | | Medium | ~500 pages/day | $5-15/month | | Heavy | ~5,000 pages/day | $50-100/month | Cost tips: - Use maxConcurrency: 1 for simple tasks - Increase concurrency for time-sensitive bulk scraping - Screenshots and full-page content use more resources --- ## Integrations ### Claude Desktop (Recommended) Best native MCP support. See Quick Start above. ### ChatGPT & Other AI Assistants Any AI supporting the Model Context Protocol can connect. Check your AI's documentation for MCP integration. ### Apify API Integrate directly into your applications: bash curl -X POST "https://api.apify.com/v2/acts/YOUR-ACTOR-ID/runs" \ -H "Authorization: Bearer YOUR-API-TOKEN" \ -H "Content-Type: application/json" \ -d '{"headlessMode": true, "maxConcurrency": 1}' ### Apify Integrations Connect scraped data to 1000+ apps: - Google Sheets - Auto-populate spreadsheets - Zapier & Make - Trigger workflows - Slack & Email - Get notifications - AWS S3, Dropbox - Export to cloud storage --- ## FAQ

What websites work with this Actor?

Almost any public website. This Actor uses a real Chrome browser with full JavaScript support—anything you can view in Chrome, the AI can automate. This includes React, Vue, Angular, and other modern frameworks. Limitations: CAPTCHAs, aggressive bot detection, phone verification requirements.

Do I need programming knowledge?

No! Describe what you want in plain English to your AI assistant. Instead of writing selectors, say "click the blue Submit button" or "extract all prices from this page".

How is this different from regular Puppeteer?

Regular Puppeteer requires writing code for every task. This Actor lets AI assistants control the browser through natural language via MCP. You describe what you want, the AI figures out how.

Can I scrape sites that require login?

Yes. Your AI can fill login forms and maintain authenticated sessions. Security note: Don't share real passwords in AI conversations—use test accounts.

What about JavaScript-heavy sites and SPAs?

Fully supported. The Actor uses a real Chrome browser with complete JavaScript execution. It handles AJAX, infinite scroll, lazy loading, and dynamic content.

How do I handle slow-loading pages?

Increase defaultTimeout in input settings (e.g., 60000ms for 1 minute). You can also ask your AI to wait explicitly: "Wait for the results to load, then extract them".

--- ## Troubleshooting ### "Element not found" errors - Ask AI to wait first: "Wait for the page to load, then click..." - Use more specific descriptions: "Click the blue 'Submit' button in the footer" - Increase defaultTimeout to 60000ms ### Timeout errors - Page takes longer than default 30 seconds - Increase timeout in Actor input - Check if site requires authentication ### "Failed to acquire browser" errors - All browsers in pool are busy - Increase maxConcurrency or wait for tasks to complete ### Connection refused - Ensure Actor is running in Apify Console - Verify endpoint URL matches the running Actor - Restart Claude Desktop after config changes --- ## Is Web Scraping Legal? Web scraping of public data is generally legal. This Actor operates ethically: - ✅ Only extracts publicly visible data - ✅ Respects robots.txt guidelines - ✅ Does not bypass authentication or CAPTCHAs - ✅ Does not collect private user data Your responsibility: Review each website's Terms of Service. For questions about your specific use case, consult legal counsel. Read more: Is web scraping legal? (Apify Blog) --- ## Privacy & Security - Isolated execution: Each run uses a sandboxed Docker container - No data persistence: Browser data deleted after each run - Non-root container: Runs as unprivileged user - Limited permissions: Actor only accesses its own storages - Session cleanup: Automatic cleanup of expired sessions --- ## Support - Issues: GitHub Issues - Apify Support: Contact Apify - MCP Documentation: modelcontextprotocol.io - Apify Docs: docs.apify.com --- ## What is MCP? The Model Context Protocol (MCP) is an open standard by Anthropic that lets AI assistants connect to external tools. Think of it as "USB for AI"—a universal way for AI models to access capabilities beyond their training. You don't need to understand MCP to use this Actor. Your AI handles all communication automatically. --- ## License MIT License - see LICENSE for details. ---

Built for the Apify community by Meysam