Skyvern

// Overview

Skyvern automates browser-based workflows using large language models and computer vision. Instead of writing custom scripts with DOM parsing and XPath selectors for each website, it uses vision LLMs to look at a page and decide which elements to click, type into, or read. This means an automation keeps working even when a site changes its layout.

It ships as a Playwright-compatible SDK that adds AI features on top of Playwright, plus a no-code workflow builder. That makes it useful for both developers who want to script automations and non-technical users who want to build them visually. You can run it locally (UI plus server) or use the managed Skyvern Cloud.

As a computer-and-browser-use agent framework, Skyvern fits tasks where one workflow needs to run across many different websites. Because it reasons about visual elements rather than relying on hard-coded selectors, the same flow can be applied to sites it has never seen before.

// What it does

• Uses vision LLMs to map on-screen elements to actions, so no custom XPath or selector code is needed per site
• Resistant to website layout changes because it reads the page visually instead of relying on fixed selectors
• Playwright-compatible SDK that adds AI capabilities on top of standard Playwright
• No-code workflow builder for non-technical users alongside the Python SDK
• Runs locally (UI + server) via pip or Docker Compose, or on managed Skyvern Cloud
• SQLite by default for the pip path, with optional Postgres via --postgres or --database-string

// Getting started

The pip path runs the full UI and server locally with a default SQLite database, so you can start without Postgres or Docker. You need Python 3.11, 3.12, or 3.13 (on Windows, also Rust and VS Code C++ build tools).

pip install "skyvern[all]"

skyvern quickstart

git clone https://github.com/skyvern-ai/skyvern.git && cd skyvern
cp .env.example .env  # add your LLM API key
docker compose up -d

Commands and code are distilled from the project's own documentation — always check the official repo for the latest.

// When to use it

• Automating a repetitive web task (form filling, data entry, status checks) across many sites with one workflow
• Extracting structured data from pages that change layout often and break selector-based scrapers
• Letting non-technical team members build browser automations through the no-code workflow builder
• Adding AI-driven element interaction to existing Playwright-based automation

// How Skyvern compares

Skyvern alongside other open-source computer & browser use tools AI/TLDR tracks, ranked by GitHub stars.