Skyvern
A tool that automates browser-based workflows using vision-language models and computer vision instead of hand-written selectors, through a Playwright-compatible SDK or a no-code workflow builder.
Skyvern automates browser-based workflows using large language models and computer vision instead of hand-written, selector-based scripts. It provides a Playwright-compatible SDK that adds AI functionality on top of Playwright, plus a no-code workflow builder so technical and non-technical users can both automate manual workflows on any website.
Rather than relying only on pre-defined XPath interactions, which break whenever a website's layout changes, Skyvern uses vision-capable LLMs and a swarm of agents to visually understand a page and plan and execute the actions needed to complete a task, following a task-driven autonomous-agent design in the style of BabyAGI and AutoGPT. This lets the same workflow run on websites it has never seen before, and keeps working through layout changes that would break a selector-based script.
It can run locally, with a UI and server backed by SQLite by default or Postgres, or through Skyvern Cloud, a managed version with anti-bot handling, a proxy network and CAPTCHA solving. It is written in Python and licensed under the AGPL-3.0 license.