Spin up an isolated sandbox with Chromium already baked in, run a scrape or test script, and pull the results back — no browser download, no local install. A headless browser API your AI agents can call in one line.
$5 free credits · no card · Chromium baked in, no browser download
$ pip install pilldrops $ python scrape_job.py from pilldrops import Sandbox with Sandbox.create(template="playwright") as sb: sb.upload("./scrape.py", "/scrape.py") r = sb.run("python /scrape.py") print(r.stdout) # scraped 128 rows sb.download("/out", "./results") # traces, pngs, json ✓ chromium baked · sandbox ready in 180ms · results in ./results
Serverless browser APIs hand you a page and nothing else. The playwright template gives you a full dedicated microVM — a real Chromium, a real filesystem for artifacts, and a public URL to watch the run live.
The playwright template ships Chromium and the Playwright runtime pre-installed. No playwright install, no download wait, no missing system libraries.
Write screenshots, PDFs, HAR files and scraped JSON anywhere on disk, then sb.download("/out","./results") pulls the whole directory back to your machine.
Launch a headed browser or a live viewer and open sb.url to watch the automation run in real time — handy for debugging flaky selectors. App preview guide →
Playwright's tracing and video recording work out of the box. Save the trace.zip to disk, download it, and open it in the Playwright trace viewer locally.
Write an ordinary Playwright script — headless by default, or headed when you want the preview URL — and run it inside the sandbox. Everything it writes to disk comes back with one download.
Load a page, pull the data you need, and dump it to /out.
# scrape.py — runs inside the sandbox from playwright.sync_api import sync_playwright with sync_playwright() as p: b = p.chromium.launch() # baked in pg = b.new_page() pg.goto("https://example.com") pg.screenshot(path="/out/home.png")
Run headed, then open sb.url to watch the form fill in real time.
with Sandbox.create( template="playwright", ports=[9222]) as sb: print(sb.url) # watch live sb.upload("./flow.py", "/flow.py") sb.run("python /flow.py").check() sb.download("/out", "./results")
The same sandbox that scrapes runs your end-to-end tests. Point pilldrops test at a suite and it drives real browsers — and Android 14 emulators — with traces and video you can download when something fails.
$ pilldrops test ./suite.sh # browser + mobile E2E → spinning up playwright template (chromium baked) ✓ login.spec 12 passed ✓ checkout.spec 8 passed ✓ android/onboarding 5 passed (Android 14 emulator) ✓ 25 passed · traces + video saved to ./results
One call spins up the playwright template with Chromium baked in, runs your scrape or test, and hands back every artifact it wrote. Per-second billing, trace + video capture, $5 free credits — no card.