Steel Cookbook
A collection of production-ready examples demonstrating how to build powerful browser automations with Steel. From basic Playwright scripts to advanced AI agent workflows, these examples show you how to leverage Steel's cloud browser infrastructure for reliable, scalable web automation.
-
Basics: Essential examples covering browser automation fundamentals with popular frameworks like Playwright, Puppeteer, and Selenium.
-
AI Agents: AI-powered browser automation using computer use models (Claude, OpenAI, Gemini) and specialized agent frameworks (Browser-use, Stagehand, CrewAI).
-
Advanced Features: Steel platform capabilities including persistent profiles, credential management, browser extensions, and session reuse.
Example catalogs
Browse examples by your preferred language:
Or explore the complete catalog below:
Basics
Foundation examples for browser automation with Steel. Start here if you're new to Steel or want to integrate it with your preferred automation framework.
| Example Name | Languages |
|---|---|
| Playwright Integration | |
| Puppeteer Integration | |
| Selenium Integration |
AI Agents
AI-powered browser automation that can understand, reason about, and interact with web pages autonomously.
Computer Use Models
Vision-based AI models that can control browsers by analyzing screenshots and executing actions.
| Example Name | Languages |
|---|---|
| Claude Computer Use | |
| Claude Computer Use (Mobile) | |
| OpenAI Computer Use | |
| Gemini Computer Use |
Agent Frameworks
Specialized frameworks that provide high-level abstractions for building AI-powered browser automation workflows.
Advanced Features
Steel platform features that enhance your browser automation workflows with persistent state, credential management, and more.
| Feature | Example |
|---|---|
| Auth Context Reuse | |
| Persistent Profiles | |
| Credential Management | |
| Browser Extensions | |
| File Handling |
Language-Specific Catalogs
TypeScript
Basics
- Playwright - Playwright with TypeScript and session management
- Puppeteer - Puppeteer with CDP connection and cloud browser automation
AI Computer Use
- Claude Computer Use - Autonomous web interactions with Claude
- Claude Computer Use (Mobile) - Claude for mobile viewports
- OpenAI Computer Use - OpenAI vision-based browser control
- Gemini Computer Use - Google Gemini computer use model
Agent Frameworks
- Stagehand - AI-powered browser automation with natural language
- Inngest AgentKit - Multi-agent networks and web scraping
- Magnitude - AI-powered testing framework
- Mastra - Typed agents with the Mastra Model Router and Studio playground
- Vercel AI SDK v6 - ToolLoopAgent with typed tools and structured output
- Vercel AI SDK v6 (Next.js) - Next.js chat app with streamText, useChat, and an embedded Live View
- OpenAI Agents SDK - Agent with tool() + Zod outputType for structured final answers
- Claude Agent SDK - Anthropic's first-party agent loop with Steel exposed as in-process MCP tools
- Deep Research - Lead orchestrator dispatches parallel researcher subagents on Claude Agent SDK, each driving its own Steel session
Advanced Features
- Auth Context - Reuse authentication across sessions
- Profiles - Persistent browser profiles
- Credentials - Credential management and injection
- Extensions - Browser extension support
- Files API - File upload and download handling
Python
Basics
- Playwright - Playwright with Python
- Selenium - Selenium WebDriver integration
AI Computer Use
- Claude Computer Use - Autonomous browser control with Claude
- OpenAI Computer Use - OpenAI computer use agent
- Gemini Computer Use - Google Gemini integration
Agent Frameworks
- Stagehand - Stagehand with Python
- Browser-use - Browser-use agent framework
- Browser-use + CAPTCHA (Auto) - Browser-use with automatic CAPTCHA solving
- Browser-use + reCAPTCHA v2 (Manual) - Manual reCAPTCHA v2 workflow
- Agno - Agno toolkit-based automation
- CrewAI - Multi-agent collaboration framework
- Notte - Notte browser agent framework
- OpenAI Agents SDK - Agent with @function_tool + Pydantic output_type
- LangGraph - Explicit state-machine agent loop with structured-output formatter
- Pydantic AI - Provider-agnostic typed agent with deps_type and output_type
- Microsoft Agent Framework - 1.0 successor to AutoGen and Semantic Kernel, plain-function
@tooldecorator - Claude Agent SDK - Anthropic's first-party agent loop with Steel exposed as in-process MCP tools
- Deep Research - Lead orchestrator dispatches parallel researcher subagents on Claude Agent SDK, each driving its own Steel session
Running the Examples
Each example contains a comprehensive README with setup instructions. Here's the general workflow:
1. Clone and Navigate
git clone https://github.com/steel-dev/steel-cookbook.git
cd steel-cookbook/examples/<example-name>
2. Install Dependencies
TypeScript:
npm install
# or
yarn install
# or
pnpm install
Python: (install uv first)
uv sync
3. Configure Environment
Copy the .env.example to .env and add your Steel API key:
cp .env.example .env
Get your API key from the Steel Dashboard.
4. Run the Example
TypeScript:
npm start
# or
npx tsx index.ts
Python:
uv run main.py
Refer to each example's README for specific requirements and advanced configuration options.
Getting Steel API Access
All examples require a Steel API key. Steel provides cloud browser infrastructure with:
- Managed browser sessions - No infrastructure setup or maintenance
- Built-in CAPTCHA solving - Automatic CAPTCHA resolution
- Proxy support - Residential and datacenter proxy integration
- Session persistence - Maintain state across multiple sessions
- Extension support - Load browser extensions in cloud browsers
Sign up for Steel to get your API key and start building.
Joining the Community
Stay connected with the Steel community:
- Discord - Get help, share feedback, and connect with other developers
- Documentation - Comprehensive guides and API reference
- X/Twitter - Stay updated with the latest news and features
This is the perfect place to ask questions, share your automations, and learn from the community!
Contributing
We welcome contributions! See the Contributing Guide for:
- Adding new examples
- Improving existing examples
- Reporting issues
- Suggesting new integrations
Support
- Documentation - Complete guides and tutorials
- API Reference - Detailed API documentation
- Discord Community - Live help and discussion