scrapybara.com

What it is

Scrapybara is a platform that provides remote desktop instances for AI computer-use agents. Its core pitch is essentially “give AI a computer.” It hosts Ubuntu, lightweight Chromium browser, and Windows desktop instances that agents can operate, and exposes a unified interface for models to complete tasks involving the web, code execution, the file system, and more. The page example shows launching an Ubuntu instance with the Python SDK and connecting OpenAI, ComputerTool, BashTool, and EditTool to carry out a task such as “scrape YC W25 companies.”

Core capabilities

From an AI perspective, Scrapybara is more agent infrastructure than an in-house foundation model product. It supports computer use models, structured outputs, multi-turn conversations, and custom tools, making it suitable for connecting model reasoning with real desktop environments. On the engineering side, the platform emphasizes launching hundreds of instances in milliseconds, performing actions with low latency, and handling orchestration and scaling. It also supports interactive stream for real-time monitoring and takeover, authenticated access for saving and loading logged-in websites, and session persistence for pausing and resuming instances. These capabilities are important for long-running automation, authenticated web workflows, and parallel tasks.

Pricing and trial

The page shows “Try for free,” “Free account,” and “Scale as you go,” indicating that there is a free account and a usage-based scaling model. However, the crawled text does not disclose the free allowance, billing unit, instance pricing, concurrency limits, or enterprise plans, so cost predictability still requires checking the documentation or confirming after signup.

Pros and cons

The main advantage is its clear positioning: it is purpose-built for real computer operation by AI agents. It provides Python and TypeScript SDKs, making integration relatively straightforward. Its instance types cover Ubuntu, browser, and Windows environments, which fit use cases such as web scraping, research, UI testing, and similar workflows. The drawbacks are that the page lacks details on data privacy and security isolation. In particular, because it supports saving access to authenticated websites, its credential management mechanisms should be carefully verified. It also does not provide Chinese-language support, SLA information, support details, or detailed pricing.

Who it’s for and access from China

Scrapybara is suitable for development teams building web-scraping agents, deep research agents, automated PR testing bots, end-to-end UI testing, or products that require large-scale remote desktops. The page does not state how well it works from China. In practice, network connectivity, overseas payments, and dependencies on external models such as OpenAI may affect usage. If access is constrained, alternatives include Browserbase, E2B, Steel.dev, or self-hosting a browser cluster with Playwright/Puppeteer.