Skyvern

Automates browser workflows with vision LLMs and a no-code builder instead of brittle DOM selectors.

Open SourceSelf Hosted
0.0 (0)

About

Instead of relying on XPath or CSS selectors that break when a page changes, Skyvern drives browsers with a swarm of agents that read pages through vision LLMs, so the same automation generalizes to websites it has never seen and survives layout changes. The Playwright-compatible SDK adds AI primitives such as page.act(), page.extract(), page.validate(), and page.prompt(), and a no-code workflow builder chains tasks with loops, conditionals, HTTP requests, and structured data extraction. It handles logins with 2FA support and password manager integrations for Bitwarden, 1Password, and LastPass, and supports OpenAI, Anthropic, Azure OpenAI, AWS Bedrock, Gemini, Ollama, and OpenRouter as model backends. Self-hosting takes pip install skyvern plus a quickstart command, or Docker Compose bundling Postgres, the API, and a UI, while a managed cloud runs at app.skyvern.com. The core is AGPL-3.0 licensed with about 22,600 GitHub stars, and the YC-backed team reports 64.4 percent accuracy on the WebBench benchmark, strongest on form-filling write tasks.

Should you use Skyvern?

Pick it when

Pick Skyvern when you must automate forms and workflows across many websites you do not control, including logins with 2FA, and want vision-based agents that survive layout changes plus a no-code builder.

Look elsewhere when

Skip it if AGPL-3.0 is a problem for a hosted product you modify, or if you automate a few known sites where cheaper deterministic steps work: Stagehand's MIT SDK mixes code and AI and caches actions.

Alternatives to Skyvern

  • Stagehand

    MIT licensed and lets you choose per step between code and AI, with cached actions for cheaper repeats; less no-code tooling than Skyvern.

  • Browser Use

    MIT Python library for browser agents with a lighter footprint; no workflow builder, password manager integrations, or bundled UI.

  • UI-TARS Desktop

    Controls native desktop apps as well as the browser through vision models; broader reach, less focused on structured web workflows.

  • Crawl4AI

    Read-only crawling to clean Markdown; far cheaper when you only need to extract content and never fill forms or log in.

Reviews (0)

Leave a Review

No reviews yet. Be the first to review!

Details

Price
Freemium
Platform
Hybrid
Difficulty
Intermediate (3/5)
License
AGPL-3.0
Added
Jul 29, 2026

Tags