Capability · Framework — agents

Skyvern

Skyvern runs a fleet of headless Chromium workers coordinated by a FastAPI server. It uses vision models to identify elements visually, survives DOM changes that break traditional scrapers, and ships an operator dashboard plus a REST API for scheduling runs.

Framework facts

Category: agents
Language: Python
License: Apache-2.0
Repository: https://github.com/Skyvern-AI/skyvern

Install

pip install skyvern
skyvern init

Quickstart

from skyvern import Skyvern

client = Skyvern()
task = client.run_task(
    prompt='Find today\'s weather in New Delhi and return temperature in C.',
    url='https://www.google.com',
)
print(task.output)

Alternatives

browser-use — lighter Python library
LaVague — research-focused action engine
Stagehand — TypeScript agent by Browserbase
Multion — commercial browser agent API

Frequently asked questions

Why pick Skyvern over browser-use?

Skyvern ships an end-to-end platform — dashboard, workers, retries, artifact storage — making it closer to a production system than a library. Pick browser-use if you want to embed agent logic inside your own app.

Which LLMs does Skyvern support?

OpenAI, Anthropic, Azure OpenAI, and Gemini via environment variables. The vision engine benefits from frontier vision models like Claude Opus or GPT-4o-class models.

Sources

Skyvern — GitHub — accessed 2026-04-20
Skyvern — docs — accessed 2026-04-20