Grid
Grid is your QA lab — a place to spin up real devices and browsers, run tests across them in parallel, and watch the results stream in live. Use it to validate that what your agents build actually works on the devices your users have. Access via Operate → Grid in the sidebar.
What it is
Grid lets you launch simulated phones, tablets, and browsers on demand, drive them with automated test suites, and inspect the results — screenshots, video, pass/fail counts — all from one screen. It’s backed by isolated sandboxes, so every device and terminal session runs in a clean, secure environment.
How it works
Pick devices from the catalog and start them; each one boots in a provisioned sandbox with a live stream you can watch in-card or fullscreen. Create a test suite, run it across your active devices, and review results in real time or browse historical runs. Grid is organized into six tabs.
Devices
Start and stop individual devices or the whole pool at once, and watch each live stream. You can:
- Choose from device models like Pixel 8, Samsung Galaxy S24, Galaxy Tab S9, and Chrome Desktop (iOS coming soon).
- Apply a network profile to simulate real-world conditions — 4G, 3G, Edge, offline, lossy, or high-latency.
- Monitor pool capacity — how many device slots are in use versus available.
Terminal
A terminal connected to an isolated devbox sandbox, pre-loaded with common runtimes (Node, Python, Go, Rust, Java, Ruby, and more) for ad-hoc commands and debugging.
Test
Create and run test suites across eight methodologies — TDD, BDD, Regression, E2E, Mobile, Website, API, and Performance. Filter suites by type, view runs with status and pass/fail counts, and drill into any individual test to see its duration, the scenario it covered, screenshots, and video.
Results
A history of every QA run, with status badges, the devices involved, and pass/fail summaries — so you can track quality over time.
Artifacts
Download the build outputs and artifacts produced during test runs.
Settings
Configure the runtime environment for your devices and sandboxes — language versions, machine specs, isolation level, network policy, secrets, snapshots, and MCP integrations.
See also: The Studio, Cendra, Cockpit