How It Works
Instead of using traditional XPath selectors that break when a site changes its layout, Skyvern uses AI agents that "see" the page and understand its contents, mapping visual elements to the required actions. It is inspired by the design of task-driven autonomous agents like BabyAGI and AutoGPT, but with the added ability to interact with websites through browser automation libraries such as Playwright.
Key Features
- AI-Powered Page Commands: Uses natural language to interact with elements (click, fill, extract, validate)
- Task & Workflow: Create single tasks or multi-step concatenated workflows
- Data Extraction: Extract structured data from web pages with JSON schema
- File Downloading: Automatically download files from websites
- Authentication Support: Supports 2FA (TOTP, email, SMS) and integrations with password managers (Bitwarden, 1Password, LastPass)
- Live Streaming: Watch in real time what Skyvern is doing on the browser
- Multi-LLM Support: OpenAI, Anthropic, Gemini, Ollama, OpenRouter, and others
Installation Options
- Skyvern Cloud: Managed version on app.skyvern.com
- Local Setup: With pip install skyvern or Docker Compose
- SDK: Python and TypeScript with Playwright-compatible APIs
Learn more at: https://github.com/Skyvern-AI/skyvern
License: AGPL-3.0 | Stars: 21.6k | Version: 1.0.36 (2026-05-10)
