Scrapybara

Scrapybara hosts remote desktop instances for computer use agents like CUA and Claude Computer Use. With its unified API, developers can write one line of code to execute agents with any model and access low-level controls like the browser, filesystem, and code sandboxes. Scrapybara handles autoscaling, authentication, and system environments, enabling users to deploy fleets of agents to production and automate any free-form computing task at scale.

How to Use

Developers can use Scrapybara by installing its Python or TypeScript SDK. After initializing a client, they can start an Ubuntu instance and then use the client.act() method with various tools (like ComputerTool, BashTool, EditTool) and a specified AI model (e.g., OpenAI) to execute prompts for automated tasks, such as web scraping.

Key Features

• Remote desktop instances for AI agents
• Unified API for agent execution with any model
• Low-level controls (browser, filesystem, code sandboxes)
• Autoscaling, authentication, and system environment management
• Instantly spin up desktop instances
• Browser, code execution, and filesystem capabilities
• Start hundreds of instances in milliseconds
• Minimal latency for computer actions
• Interactive stream to monitor and yield control
• Authenticated access to save and load websites
• Session persistence to pause and resume instances

Use Cases

• Deploy intelligent agents to automate free-form computing tasks at scale
• CodeCapy.AI: PR bot for end-to-end UI testing in Ubuntu instances
• CopyCapy: Scrape and 'capyfy' any website
• Wide Research: Deep research scraping in parallel
• Dungeon Crawler: Fight monsters and explore dungeons