Addresses that are not public
The URLs of staging and pre-production environments, and which of your services answer, in what order.
Your tests stay Playwright, Selenium and pytest. We don't replace them. We make them enterprise-ready, and you own everything: code, evidence and history.
Every run records your app as it really is. Choose where that record lives, and what it costs you.
Automation sees what your users see, and it writes it down. That record is the point of testing, and it is also sensitive.
The URLs of staging and pre-production environments, and which of your services answer, in what order.
The accounts and passwords tests sign in with. A browser test recorded in the Studio keeps the password you typed, so the file is as sensitive as the account.
For each failed run, one screenshot of the screen the test last saw and, for a real failure, a video. They are kept exactly as captured.
Network captures such as HAR files and traces record headers and cookies unless they are scrubbed. The Studio writes no HAR files, masks secret headers and cookies in API evidence, and keeps a trace as captured when you ask for one. Console errors and stack traces describe how the app is built.
None of this makes a provider careless. It means the controls are theirs, and your policy has to accept that.
The provider operates the machine your runs execute on and the disks the evidence is written to.
Copies follow the provider's backup policy, on the provider's schedule, in places you do not list.
Provider staff and contractors have administrative paths to the hardware, under the provider's controls rather than yours.
Requests made under the law of the provider's jurisdiction can reach data on the provider's hardware, sometimes without notice to you.
When you cancel, deletion follows the provider's retention schedule, not the day you decide.
Your workloads can share physical machines with other customers, separated by software.
Pick one to see where the evidence sits and what crosses the line out of your network. The table compares all four.
Your network
Your browser, on the vendor's websiteNexus Studio on your PCNexus Studio on the mini PCNexus Studio on your PC The app you testTest runs and evidenceAndroid phones on USBRuns while it sleepsSomeone else's data center
Test runs, screenshots, video and logsDevice farm phonesTest runs and evidence, until you pull themAndroid phones on USBLicense check: a machine hash, a nonce and a token| Question | Cloud QA platform | Rented VPS | Mini PC you own | Your own PC |
|---|---|---|---|---|
| Where the evidence lives | On the vendor's storage | On the provider's disks, until you pull it home | On a disk in your office | On your own disk |
| Who can reach it | The vendor's staff and systems, under the vendor's policies | The provider, through the hypervisor and backups, and anyone with your SSH key | You, and whoever you let onto that machine | You |
| What leaves your network in a run | The whole run: your app's screens, traffic and logs | The run itself; results stay there until pulled | The license check: a machine hash, a nonce and a token; the key only once, at activation | The license check: a machine hash, a nonce and a token; the key only once, at activation |
| Cost pattern | A recurring subscription set by the vendor | Monthly rent | One purchase, plus power | Nothing new |
| Real Android phones on USB | The vendor's device farm, if offered | No: a rented server cannot reach a phone on your desk | Yes, plugged into it | Yes, plugged into it |
| Runs around the clock | Yes | From a CI schedule; Watch runs on the Studio's own machine | Yes: it stays on, so Watch keeps its schedule | Only while it is awake |
A mini PC is optional, and recommended: a dedicated machine that stays on, for nightly regressions and scheduled runs without tying up a workstation. Each workspace picks where the engine runs its suites, on the Studio's Docker screen: this machine, Docker on this machine, or a VPS over SSH running the same container image. None of them is required, and switching is one setting. The Studio itself is a desktop app and never runs in Docker or over SSH.
Install the Studio on Windows 11 and run there; nothing else is required. Phones plug into it, and scheduled runs happen while it is awake.
The same Studio on a dedicated Windows 11 machine that stays on, for nightly regressions and scheduled runs without tying up a workstation. Phones plug into it, Watch keeps its schedule, and the evidence stays on its disk. Or the box runs Ubuntu 24.04 LTS with Docker Engine and Compose v2: the Studio on your PC runs the suites on it over SSH, like a VPS on your own network, and phones stay on the Studio's PC. Either way it counts as one machine of your plan.
Set per workspace. The Studio stays where it is; the suites run inside the runner container on the same machine and disk, so each run starts from the same image. Phone runs stay on the machine itself.
Set per workspace. A VPS with Ubuntu 24.04 LTS, Docker Engine and Compose v2, reached with SSH key authentication, runs the suites in the same container image, and the evidence stays there until you pull it. Phones, Selenium and Python suites and infrastructure chaos stay on the Studio's machine. It makes sense for extra capacity, or next to a staging environment already in a data center, when your policy allows evidence on rented disks.
What to look for
Windows 11 x64, or Ubuntu 24.04 LTS for a box that only runs the engine. 16 GB of RAM is the minimum; 32 GB is recommended, for steadier runs when several tests open their browsers in parallel. Three browsers at once need about 2.5 GB free by the engine's own estimate, on top of the system and the Studio.
An SSD for the evidence and the videos, and free USB ports for the phones.
These are recommendations with their reasons, not benchmarks. Disk space and what the installer bundles: system requirements.
Compare a monthly bill for a cloud QA platform or a rented server with a mini PC you buy once. This compares hardware and power with a cloud subscription for one parallel session. Your Nexus Studio plan, priced per machine and never per test, applies either way, so it is left out, and so is the time you spend looking after the box.
Example values. Replace them with your own.
Break-even:
After 1 year:
After 3 years:
Over the months you picked:
Where the example comes from, read on 2026-09-30. Sauce Labs Virtual Device Cloud: USD 149 a month billed annually (USD 199 month to month) for 1 parallel test, USD 1,788 a year (saucelabs.com/pricing). BrowserStack Automate, Desktop & Mobile: USD 175 a month billed annually, USD 2,100 a year, its entry option (browserstack.com/pricing). Mini PC prices read on 2026-10-06, retail, x86-64 with a 500 GB SSD or more: about USD 340 to 385 with 16 GB of RAM, about USD 530 to 610 with 32 GB, the size we recommend. Power: USD 5 a month, an example.
Estimate. In year one, a mini PC of about USD 350 plus about USD 60 of power costs about USD 1,400 to 1,700 less than one year of either subscription. Each year after, the difference is about USD 1,700 to 2,000.
Four questions. The answer comes with its reasons, and the four setups below say when each one fits.
Answer the four questions, or read the four setups below: each one says when it fits.
Fits one person without a schedule. Phones plug into it and the evidence stays on your disk; runs pause while it sleeps.
Optional. Fits scheduled runs and teams, with or without phones or strict evidence rules. It stays on, holds the evidence, and costs one purchase.
Add it when you want every run to start from the same image. Phone runs stay on the machine itself.
Fits scheduled runs without phones, when your policy allows evidence on rented disks. It runs the same container image as Docker. Pull the evidence home after each run, and schedule it from CI: Watch runs on the Studio's own machine.
Free for 30 days, with every feature, on your PC or a mini PC. No card needed.