Multi-language browser automation plugin enabling AI agents to browse websites, interact with elements, and extract data. Available in TypeScript, Python, and Rust with full feature parity.

plugin-browser/
├── protocol/           # Shared protocol definitions (JSON Schema)
├── typescript/         # TypeScript/Node.js implementation
├── python/            # Python implementation
├── rust/              # Rust implementation
└── README.md          # This file

All implementations support:

• Navigation: Navigate to URLs, go back/forward, refresh pages
• AI-Powered Interactions: Click, type, and select elements using natural language
• Data Extraction: Extract structured data from web pages
• Screenshots: Capture page screenshots
• CAPTCHA Solving: Automatic CAPTCHA solving (Turnstile, reCAPTCHA, hCaptcha)
• Session Management: Handle multiple browser sessions
• Security: URL validation, domain filtering, rate limiting
• Retry Logic: Exponential backoff for reliability

cd typescript
npm install
npm run build

import { browserPlugin } from "@elizaos/plugin-browser";

const agent = {
  plugins: [browserPlugin],
};

cd python
pip install -e .

from elizaos_browser import create_browser_plugin

plugin = create_browser_plugin()
await plugin.init()
await plugin.handle_action("BROWSER_NAVIGATE", "Go to google.com")

cd rust
cargo build

use elizaos_browser::create_browser_plugin;

let mut plugin = create_browser_plugin(None);
plugin.init().await?;
plugin.handle_action("BROWSER_NAVIGATE", "Go to google.com").await?;

All implementations support these actions:

All implementations use the same environment variables:

# Browser settings
BROWSER_HEADLESS=true
BROWSER_ENABLED=true
BROWSER_SERVER_PORT=3456

# Cloud browser (optional)
BROWSERBASE_API_KEY=your_api_key
BROWSERBASE_PROJECT_ID=your_project_id

# AI providers for intelligent interactions (optional)
OPENAI_API_KEY=your_openai_key
ANTHROPIC_API_KEY=your_anthropic_key
OLLAMA_BASE_URL=http://localhost:11434
OLLAMA_MODEL=llama3.2-vision

# CAPTCHA solving (optional)
CAPSOLVER_API_KEY=your_capsolver_key

The protocol/ directory contains:

• schema.json: JSON Schema defining all types, actions, and messages
• README.md: Protocol documentation

All implementations communicate with the browser server using WebSocket and follow the same message protocol, ensuring interoperability.

All implementations use consistent error codes:

• URL validation with domain allowlists/blocklists
• Input sanitization to prevent XSS/injection
• Rate limiting for actions and sessions
• Protocol restrictions (HTTP/HTTPS only by default)

cd typescript
npm install
npm run build
npm test
npm run typecheck

cd python
pip install -e ".[dev]"
pytest
mypy elizaos_browser
ruff check elizaos_browser

cd rust
cargo build
cargo test
cargo clippy
cargo fmt

MIT

Built with:

• Stagehand - AI-first browser automation framework
• Playwright - Cross-browser automation
• CapSolver - CAPTCHA solving service