Skyvern Advanced Browser Automation

What is Skyvern?

Skyvern automates browser-based workflows using LLMs and computer vision, providing an API endpoint to replace brittle or unreliable browser automation solutions. It is designed for anyone needing to automate manual workflows across many websites without custom scripts.

How to use Skyvern?

Install via pip install skyvern, then run skyvern quickstart to launch the server and UI locally. Tasks can be executed using the Python SDK with skyvern.local.run_task() or via Skyvern Cloud with an API key. Docker Compose and managed cloud options are also available.

Key features of Skyvern

• Tasks: single browser automation requests with URL and prompt.
• Workflows: chain multiple tasks for complex processes.
• Livestreaming: view the browser viewport in real time.
• Form filling and data extraction with optional schema.
• File downloading with automatic block storage upload.
• Authentication support including 2FA (TOTP, email, SMS) and password managers (Bitwarden).
• Model Context Protocol (MCP) support for LLM integration.
• Integrations with Zapier, Make.com, and N8N.

Use cases of Skyvern

• Downloading invoices from multiple supplier websites.
• Automating job application submissions across different portals.
• Procuring materials by navigating e‑commerce and supplier sites.
• Registering accounts and filling forms on government websites.

FAQ from Skyvern

What runtime does Skyvern require?

Skyvern requires at least Python 3.11 to run locally.

Can I run Skyvern without managing infrastructure?

Yes, Skyvern Cloud is a managed version that handles infrastructure, anti‑bot detection, proxies, and CAPTCHA solvers.

Does Skyvern support two‑factor authentication?

Yes, it supports QR‑based (Google Authenticator, Authy), email‑based, and SMS‑based 2FA.

How does Skyvern handle website layout changes?

It uses visual LLMs rather than fixed XPaths, making it resistant to layout changes and able to work on unseen websites.

Does Skyvern support the Model Context Protocol?

Yes, Skyvern supports MCP to allow any LLM that supports MCP to control the browser. See the separate MCP documentation for details.