AI-powered browser automation for websites, with vision-based workflows and a Playwright-compatible SDK
- Stars21.9k
- Forks2k
- Open Issues164
AGPL-3.0
- Python
- TypeScript
- Jinja

About Skyvern
Skyvern automates browser-based workflows using LLMs and computer vision. It helps replace brittle script-driven automation for tasks like logins, form fills, and data extraction across websites that change layout often.
It provides a Playwright-compatible SDK that adds AI actions on top of Playwright, plus a no-code workflow builder. Skyvern uses vision LLMs and a swarm of agents to map visual elements to actions, so a workflow can run on sites it has not seen before and fall back from selectors to AI when needed.
Skyvern offers a managed cloud service and a local setup. The open source core is licensed under AGPL-3.0, telemetry is on by default but can be turned off, and the cloud version adds anti-bot measures, proxy network, and CAPTCHA solvers.
Key features
- Playwright-compatible SDK with AI actions
- No-code workflow builder
- Vision-based browser interaction
- Selector-first with AI fallback
- Data extraction from web pages
Details
- First released
- 2024
- Telemetry
- On by default, opt out via SKYVERN_TELEMETRY
- License
- AGPL-3.0
- Self-hosting
- Local setup with UI and server
- Cloud
- Managed cloud available
- CLI
- Windows · WSL · macOS · Linux
