Back to full roadmap
topicadvanced
Browser-Use (Open Source)
Open-source browser agent — Playwright + LLM. Self-host, vendor-agnostic.
3 hours2 resources1 prereqs
Browser-Use: Python library, Playwright + any LLM. Uses a DOM parsing + vision hybrid approach. Cheaper than Anthropic/OpenAI computer use because it extracts DOM (not screenshot every step) + spot-checks vision.
Stack alternatives:
- Stagehand (Browserbase) — TS-based, structured action API
- Skyvern — vision-first, low-code workflow builder
- AgentE — Microsoft (HAGENT), new on GitHub
- Native Playwright + LLM — max control, most flexible