Browser-use hit 31,000 stars (CuratedRepo, Apr 2026) by making modern SPAs accessible to AI agents via Playwright + vision models — solves complex sites, captchas, autonomous browser workflows. From Junagadh I use it where n8n HTTP fails: JS-heavy portals, OTP flows, anti-bot checkouts.
I curate at /repos — tag browser-agent. Contact if your portal blocks plain HTTP automation.
Why Browser-Use Now
Fungies trends Apr 2026: future is not one agent doing everything — teams of specialized agents collaborating. Browser-use is the "eyes" specialist alongside RAG and workflow agents. OSSInsight trending category: Browser agents in top 4 lanes April 2026.
| Need | Use n8n HTTP | Use browser-use |
|---|---|---|
| API exists | ✅ Fast | Overkill |
| SPA / JS-rendered / captcha | ❌ Fails | ✅ Playwright + vision solves |
| Cost | Low | Higher vision tokens |
It gives agents "unmetered search" alternative to SERP APIs, like SearXNG 17K does for search grounding but for interaction.
from browser_use import Agent
agent = Agent(task="Login to GST portal, download GSTR-2A, save to /data", llm=llm)
await agent.run()
Same ledger: every browser action = OTel step with screenshot hash for audit. HITL before any irreversible (pay, submit).
Bottom Line: Browser-use 31K in 2026 fills the gap HTTP automation cannot — JS SPAs, captchas, portals. Pair it with n8n: n8n orchestrates, browser-use eyes the page.
Frequently Asked Questions
Does it solve captchas?
Vision + Playwright handles many, but HITL pause for sensitive captchas — I enforce review before submit.
Is it heavy?
More tokens than API, but cheaper than human 4 hours. Hybrid: API if exists else browser.
Alternative?
Playwright alone = code; browser-use = agent reasoning over Playwright. Agent loops.
Who automates portals from Junagadh?
Deepak Bagada — browser-use + Pydantic + OPA, shadow mode first 30 days.