Some links on this page are affiliate links: if you buy through them we may earn a commission, at no extra cost to you.
Run Playwright in the cloud in one of two ways: install Playwright and its browsers in a cloud CI runner you control, or keep your tests in CI and connect them to a managed browser service such as Azure Playwright Workspaces or BrowserStack Automate. The first approach gives you infrastructure control and predictable tooling. The second offloads browser hosting, matrices and some scaling. Start with self-managed CI unless you specifically need remote browser/device coverage, private-network tunnels or provider-hosted artifacts.
Choose the cloud operating model first
Your test code can stay almost identical in either model. What changes is where browser processes run and who maintains them.
Self-managed cloud CI
A GitHub Actions, GitLab, Azure DevOps or similar runner installs your project dependencies, Playwright’s matching browser binaries and (on Linux) required operating-system packages. Tests execute inside that runner. You own the image, caching, concurrency, network route and artifact retention.
Managed cloud browsers
Your CI job still starts the Playwright command, but the browser runs in a provider’s service. You configure a workspace or provider capability set, authenticate with a secret or identity, and select browsers, operating systems or devices supported by that account. The provider operates the browser fleet and normally supplies a dashboard and run artifacts.
#1 Best Overall
| Decision | Self-managed runner | Managed service |
|---|---|---|
| Browser maintenance | Your image and dependency pipeline | Provider fleet; you follow its supported versions |
| Parallelism | Runner capacity and CI jobs | Account and plan limits; verify before sizing |
| Private applications | Runner must have network access | May require a vendor tunnel, such as BrowserStack Local |
| Artifacts | Configure CI reports, traces and videos | Provider dashboard and documented logs/video may be available |
| Cost model | Cloud-runner minutes and storage | Provider-specific; Azure documents billing by total test minutes |
Do not assume a provider’s example worker count, browser list or “10x” parallel-speed statement is an entitlement or independent benchmark. Check the current plan, region and supported-combinations pages for your account.
Run Playwright in your own cloud CI runner
1. Pin dependencies
Commit your lockfile and use a reproducible Node.js version in the CI image. A dependency upgrade can require a new Playwright browser build, so treat the package and browser installation as one change.
2. Install browsers and Linux dependencies
For a Node project, the documented baseline is:
npm ci
npx playwright install --with-deps--with-deps installs the operating-system packages needed by supported Linux runners. If your image already contains those packages, npx playwright install is sufficient. The official Playwright Docker image is another documented option for Linux agents; pin the image version rather than silently floating to a new browser build.
Recommended Free Tools
3. Execute tests
npx playwright testStore the HTML report, traces, screenshots and videos as CI artifacts when a test fails. Keep secrets out of test source; inject base URLs, credentials and tokens through the CI secret store.
4. Start with one worker
Playwright’s CI guidance recommends setting workers to 1 “to prioritize stability and reproducibility.” This is guidance, not a performance measurement. After observing CPU, memory, browser crashes and test isolation, raise workers deliberately or shard the suite across separate CI jobs.
// playwright.config.ts
import { defineConfig } from '@playwright/test';
export default defineConfig({
workers: process.env.CI ? 1 : undefined,
retries: process.env.CI ? 2 : 0,
reporter: process.env.CI ? [['html', { open: 'never' }], ['line']] : 'list',
use: {
baseURL: process.env.BASE_URL,
trace: 'retain-on-failure'
}
});5. Add parallelism safely
Use more workers only when the runner has spare resources and tests do not share mutable state. For large suites, shard by CI job instead of putting an unlimited worker count on one machine. Give each shard isolated test data, ports and output directories.
Rank #2
Connect tests to Azure Playwright Workspaces
Microsoft’s current product name is Playwright Workspaces, available through Azure App Testing. The former Microsoft Playwright Testing service was scheduled to retire on March 8, 2026; do not create new setup instructions around that retired name.
Provision and configure
- Create a Playwright Workspace in Azure App Testing in the region appropriate for your application and compliance requirements.
- Follow the current quickstart to obtain the region-specific service endpoint and add the service package/configuration to your Playwright project.
- Choose the Playwright Test Runner integration or a CDP connection as your workflow requires.
- Authenticate from CI. Microsoft’s quickstart supports Microsoft Entra ID and access tokens and strongly recommends Entra ID. Treat an access token like a long-lived password: keep it in CI secrets, limit its scope and rotate it according to your policy.
- Run a small representative suite before enabling broad parallelism. Azure documents billing by total test minutes, so measure consumption before estimating a recurring budget.
Microsoft’s example uses 20 workers. That is an example configuration, not a universal quota or promise of 20-way concurrency. Verify current quotas, package limits, regions and rate-card details for your subscription.
Network and data checks
- Confirm that the application endpoint is reachable from the Workspace service, not merely from your laptop.
- Decide how authentication, test data, certificates and allow-lists work from the service’s network.
- Review artifact retention, data residency and security settings for the selected region.
Run on BrowserStack Automate
BrowserStack Automate documents Playwright execution on cloud browsers and devices, CI integration, parallel runs and artifacts such as logs and video.
- Use BrowserStack’s current Playwright setup and capabilities documentation to select an exact browser, operating system or device combination supported by your plan.
- Add the required BrowserStack credentials as CI secrets and configure the provider’s Playwright connection in your project.
- Set the CI command and artifact policy, then run one small suite to verify authentication, capabilities and timing.
- For an internal application, configure BrowserStack Local to establish the tunnel that lets cloud browsers reach the private network. Verify routing, DNS, certificates and allow-lists rather than assuming the tunnel fixes every network issue.
Confirm current Playwright versions, browser/OS/device combinations, concurrency and commercial terms before committing a large matrix. Provider claims about parallel speed are vendor statements, not independently comparable benchmarks.
Browser versions, channels and upgrades
Each Playwright release expects matching browser binaries. After upgrading Playwright, rerun browser installation in the image or job cache:
Do these 3 things before closing this tab:
1Clear out junk files and repair common Windows errors2Fix the driver behind crashes, sound loss and screen glitches3Repair Windows errors before they cause bigger problemsnpm ci
npx playwright install --with-deps
npx playwright testPlaywright supports Chromium, WebKit and Firefox. It can also use branded Chrome and Edge channels when those browsers are installed and configured on the runner. A managed provider may expose a different set of branded channels; use its compatibility table instead of assuming parity with local Playwright.
Rank #3
Cache carefully
Cache the package manager’s downloads and Playwright browser directory only when the cache key includes the lockfile and Playwright version. A stale browser cache can produce launch errors after an upgrade; deleting the cache and reinstalling is a faster recovery than debugging a mismatched binary.
Private applications and identity
Classify the application before selecting a service:
- Public: the runner or provider can resolve and reach the URL directly.
- Runner-reachable: your self-managed CI network can reach it, but a hosted browser cannot without peering, an allow-list or a tunnel.
- Private-only: plan a supported tunnel or an approved network integration before writing tests.
Use short-lived identity where possible. Put API keys, Entra credentials, access tokens and application passwords in the CI secret manager; never commit them to playwright.config, traces or test fixtures. Scrub secrets from videos, logs and failure screenshots.
Reliability, speed and cost planning
Make failures diagnosable
- Capture a trace on the first retry or on failure.
- Upload the HTML report and relevant screenshots/videos as job artifacts.
- Log the Playwright version, browser channel, worker count, shard index and target URL.
- Use deterministic test data and isolate accounts between parallel jobs.
Estimate capacity from a real run
Measure a representative suite with one worker, then record wall-clock time, total browser time, memory and failure rate. Increase workers or shards in small steps. For a managed service, estimate billable test minutes from actual runs and check whether waiting, retries and parallel sessions are included in the provider’s current definition.
Control recurring spend
Run pull-request smoke tests on every change and reserve the full cross-browser/device matrix for scheduled or release workflows. Set CI timeouts, cap retries and stop orphaned jobs. Azure’s documentation specifically advises beginning with a small run because usage is billed in total test minutes; other providers have their own metering.
Troubleshooting common cloud failures
“Executable doesn’t exist” or browser launch failure
Cause: browsers were not installed, or the cache contains binaries for another Playwright version. Fix: run npx playwright install --with-deps, include the Playwright version in the cache key and rebuild the image.
Tests pass locally but time out in CI
Cause: slower CPU, missing OS packages, DNS differences, blocked outbound traffic or an incorrect base URL. Fix: print the resolved URL, test DNS and HTTPS from the runner, install dependencies, and replace fixed sleeps with locator or network assertions.
PC Slower Than It Used to Be?
A free scan shows the junk files, broken settings and background clutter dragging Windows down - then fixes them in one click.Free scan · Windows 10 & 11Crashes, No Sound, or Screen Glitches?
Random freezes, missing sound and display glitches usually trace back to one bad driver. Find and replace yours safely.Free scan · under a minuteManaged browser cannot open an internal URL
Cause: the hosted browser is outside your private network. Fix: use BrowserStack Local or the selected provider’s approved network path, then verify firewall rules, DNS, certificates and application allow-lists.
Authentication or endpoint errors
Cause: an expired token, wrong region endpoint, missing CI secret or malformed provider configuration. Fix: rotate the secret, validate the endpoint from the current quickstart, mask values in logs and run a one-test smoke job.
Flaky failures after enabling workers
Cause: shared accounts, ports, files or test data. Fix: return to one worker, isolate state per test or shard, and increase concurrency only after the suite is deterministic.
Unexpected cloud cost
Cause: retries, broad matrices, long waits or orphaned jobs. Fix: set job and test timeouts, cancel superseded workflows, measure a small suite and confirm the provider’s current billing definition and quotas.
What’s actually slowing this PC down?
Pick the symptom - the matching free tool is one click away.
Or skip the browser setup
If your goal is a clean image or PDF rather than an interactive test, ScreenshotNeo returns the result from one request. It accepts cookie/consent banners before capture and removes more than 60 known consent platforms, newsletter popups and chat widgets; each cleanup step can be disabled. Bot checks, blank pages, failed loads, timeouts and cache hits are not billed, and response headers identify the page verdict and billing status. Its MCP server provides take_screenshot, get_page_info and capture_pdf tools for Claude, Cursor and other MCP clients.
Using the ScreenshotNeo API documentation:
curl -G "https://api.screenshotneo.com/v1/shot" -d access_key=YOUR_API_KEY --data-urlencode url=https://stripe.com -o shot.webpimport requests
r = requests.get("https://api.screenshotneo.com/v1/shot", params={"access_key": "YOUR_API_KEY", "url": "https://stripe.com"}, timeout=90)
open("shot.webp", "wb").write(r.content)const q = new URLSearchParams({ access_key: 'YOUR_API_KEY', url: 'https://stripe.com' });
const res = await fetch(`https://api.screenshotneo.com/v1/shot?${q}`);There is a free allowance of 1,000 screenshots per month with no card. Paid plans start at $5 for 3,000 shots; every feature is included on every plan. Create a free ScreenshotNeo account to try it.
Best Value
FAQ
Can I run Playwright in a container?
Yes. Use a pinned Playwright Docker image or install the project browsers and Linux dependencies in your own container image, then run the same CI command.
Should every CI job use one worker?
One worker is the recommended starting point for CI stability. Increase workers or shard only after measuring resource use and confirming tests isolate their state.
Is Azure Playwright Workspaces the same as the retired Microsoft Playwright Testing service?
No. Microsoft directed users from the former service, retired March 8, 2026, to Playwright Workspaces in Azure App Testing. Follow the current Workspace documentation for setup and billing.
Frequently Asked Questions
Can I run Playwright in a container?
Yes. Use a pinned Playwright Docker image or install the project browsers and Linux dependencies in your own container image, then run the same CI command.
Should every CI job use one worker?
One worker is the recommended starting point for CI stability. Increase workers or shard only after measuring resource use and confirming tests isolate their state.
Is Azure Playwright Workspaces the same as the retired Microsoft Playwright Testing service?
No. Microsoft directed users from the former service, retired March 8, 2026, to Playwright Workspaces in Azure App Testing. Follow the current Workspace documentation for setup and billing.
Quick wins for a faster PC:
Repair Windows errors before they cause bigger problemsFix Now →Scan for outdated or missing drivers - takes under a minuteDriver Scan →Clear out junk files and repair common Windows errorsFree Scan →Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
