Stagehand V2: FULLY FREE Browser Use AI Agent! Automate ANYTHING! (Opensource)

Key Concepts:

• AI Web Browser Agents
• Browser Automation
• Headless Browsers
• Stage Hand (Open-Source Framework)
• MCP (Marketplace) Integration
• Natural Language Control
• AI Model Evaluation (Accuracy, Cost, Speed)
• Browserbase (Serverless Platform)
• Client/Cursor (Coding Agents)
• Large Language Models (LLMs)
• Web Scraping

1. Introduction to Browser Automation and the Problem with Existing Solutions

• AI web browser agents can automate almost any task on the web.
• Existing solutions are often locked behind paywalls (e.g., OpenAI's Computer Operator at $200/month).
• Browserbase offers a serverless platform for running, managing, and monitoring headless browsers at scale.

2. Stage Hand: An Open-Source Framework for Robust Browser Automation

• Stage Hand is an open-source framework developed by Browserbase for building robust browser automations with AI.
• It allows controlling web browsers using natural language, code, and AI tools.
• Designed to be reputable, resilient, and customizable, addressing shortcomings of traditional tools (Playwright, Puppeteer, Selenium) and agent-only solutions.
• Stage Hand 2.0 introduces new agents for seamless integration with advanced AI models for complex multi-step browser workflows.

3. Stage Hand's Capabilities and Use Cases

• Example: AI agent automatically applying for jobs, filling out forms, uploading resumes, and navigating job portals.
• Cost-effective: Approximately 10 cents per hour.
• Automates manual tests, saving hours of work.

4. Setting Up Stage Hand with Client/Cursor (MCP Integration)

• Easier installation through MCPs like Client or Cursor compared to local installation from GitHub.
• Prerequisites:
  • VS Code installed
  • Client extension installed (free)
  • Browserbase account with API key and project ID
  • LLM API key (OpenAI, local model via Ollama, etc.)
  • Node.js and npm installed
• Step-by-step installation process using Client:
  1. Open VS Code and access the Client extension.
  2. Provide API key for the desired model provider in the settings tab.
  3. Enable MCP marketplace in the features tab.
  4. Search for "Stage Hand" in the marketplace and install the MCP.
  5. Client autonomously sets up the MCP server.
  6. Provide API key for OpenAI or a local model (e.g., Ollama) when prompted by Client.
  7. Set up Browserbase keys by creating an account and copying the project ID and API key from the dashboard.

5. Using Stage Hand with Client: Commands and Examples

• Stage Hand commands available within Client:
  • stage hand navigate: Navigates to a specified URL.
  • act: Performs actions on a web page.
  • extract: Extracts data from a web page.
  • observe: Visualizes what is happening on the web page.
• Example 1: Searching for the latest React documentation and extracting main concepts:
  • Prompt: "Can you search for the latest React documentations and extract the main concepts?"
  • Stage Hand uses stage hand navigate to go to react.dev.
  • It then extracts the main React concepts and displays them in an MD file.
• Example 2: Scraping the 10 most popular videos from a YouTube channel:
  • Prompt: "I want you to head over to my YouTube channel and scrape the 10 most popular videos, extract the title, the view count, and the description for me and use the stage hand MCP to execute this task."
  • Stage Hand navigates to the specified YouTube channel URL.
  • It filters the videos by popularity.
  • It extracts the title and view count of the top 10 videos.
  • The description extraction failed due to the prompt not specifying to click on each video individually.
  • A live preview of the AI agent's actions is available on Browserbase.

6. AI Model Evaluation and Recommendations

• Stage Hand includes a model evaluation feature comparing AI model performance across accuracy, cost, and speed.
• Recommended model: Deep Seek R1 distilled Quen 32B (fastest and cheapest).
• Most accurate model: Gemini 2.0 Flash (due to multi-modal capabilities).

7. Browserbase Session Logs

• A full log of all actions performed by Stage Hand is available on Browserbase's session page.

8. Conclusion

• Stage Hand AI is a revolutionary open-source AI agent designed for seamless browser automation and web scraping.
• It offers a cost-effective and customizable solution compared to existing paywalled options.
• The video encourages viewers to explore Stage Hand using the links provided in the description.