How to test your flow
How to test your flow in Build's Test view — run the auto-derived test suite, try a Deep test, watch the Signals your flow emits, and review the Issue inbox for problems found in your deployed flow.
Once you've built a flow, the Test view helps you make sure it actually works — before it reaches your stations. This guide covers how to test your flow: running the test suite, using Deep test, watching the Signals your flow emits, and working through the Issue inbox.
Switch to Test from the tabs in the top bar. Build derives a test suite from your flow, so you'll want a built flow first.
Run the test suite
The Test suite is generated automatically from your flow and grouped into categories — Hardware, Payments, Scenarios, Hallucination, and Regression. You can:
Run all to run the whole suite, or run a single category.
See each test's last result (passed / total), and Stop a run in progress.
Update tests when you see the stale warning ("Your app changed since the tests were written") — this regenerates the suite to match your current flow.
Inspect a test
Select a test to open the inspector, which has Steps, Signals, and History tabs. A test runs in one of three modes:
auto — the harness performs the steps (navigate, click, type) and checks the result for you.
manual — you perform the actions in the preview while the harness listens; the step passes the moment the expected signal arrives, or you can mark It worked / It failed.
agent — an AI drives the preview through the steps.
If a test's selector goes stale because your flow changed, Build can self-repair it.
Try a Deep test
Deep test lets you check a whole scenario without scripting it. Type what you want to happen in plain language — for example, "Sell 3 items and print the receipt" — and select Go. An AI agent drives your flow to attempt the scenario and reports how it went.
Watch the Signals
Your flow talks to the point of sale by sending commands. The Signals tab shows every command your flow emits, live — poke the preview and watch them appear. This is how tests confirm a step worked (they wait for the expected signal). You can clear signals or approve the current set as a baseline to compare future runs against.
Review the Issue inbox
The Issue inbox collects problems found in your deployed flow. To feed it, turn on Enable debug features when you deploy — that streams runtime errors and performance issues from the live build into Test.
Issues are sorted into New, Open, and Resolved, each tagged by kind — Error, Perf, or Optimize — along with the version it appeared in and how many times it's been seen. For each issue you can:
Resolve it or Mark fixed once you've addressed it.
Promote to test — turn the issue into a test so the problem is caught automatically if it ever comes back.
Tips
Run the suite after any significant change, and use Update tests whenever the stale warning appears so your tests match your flow.
Reach for Deep test when you want to check an end-to-end scenario the way a real cashier would run it.
Step by step
- 1
Open the Test view
In the top bar, switch to Test. Build derives a test suite from your flow, so you'll need a built flow first — an empty flow shows "Build something first."
- 2
Run the test suite
Select Run all to run everything, or run a single category — Hardware, Payments, Scenarios, Hallucination, or Regression. Each test shows its last result (passed/total), and you can Stop a run in progress. If your flow changed since the tests were written, you'll see a stale warning — choose Update tests to regenerate them.
- 3
Inspect a test
Pick a test to open the inspector, with Steps, Signals, and History. A test runs in one of three modes: auto (the harness performs the steps and checks the result), manual (you perform the actions while the harness listens — it passes the moment the expected signal arrives, or you mark It worked / It failed), or agent (an AI drives the preview).
- 4
Try a Deep test
In the Deep test box, describe a scenario in plain language — for example, "Sell 3 items and print the receipt" — and select Go. An AI agent drives your flow to attempt it and reports how it went.
- 5
Watch the Signals
Open the Signals tab to see every command your flow sends to the point of sale, live. Poke the preview and watch them appear — this is how tests confirm a step worked. You can clear signals or approve the current set as a baseline.
- 6
Review the Issue inbox
The Issue inbox collects problems found in your deployed flow — turn on "Enable debug features" when you deploy to stream runtime errors and performance issues into Test. Issues are sorted into New, Open, and Resolved, each tagged by kind (Error, Perf, or Optimize) with the version and how often it was seen. Resolve or mark an issue fixed, or Promote it to a test to guard against it coming back.
Frequently asked questions
Where do I test my flow?
In the Test view — switch to it from the tabs in the top bar of the workspace.
Where do the tests come from?
Build derives the test suite automatically from your flow. When your flow changes, a stale warning appears and you can choose Update tests to regenerate them.
What is a Deep test?
A Deep test lets you describe a scenario in plain language, like "Sell 3 items and print the receipt." An AI agent then drives your flow to try it and reports the result.
What are Signals?
Signals are the commands your flow sends to the point of sale, shown live as they happen. Tests confirm a step by watching for the expected signal.
What's in the Issue inbox?
Runtime errors and performance issues from your deployed flow, streamed into Test when you enable debug features at deploy. You can resolve them, mark them fixed, or promote one to a test.
What do auto, manual, and agent mean?
Auto means the harness runs the test's steps and checks the result. Manual means you perform the actions while the harness listens for the expected signal. Agent means an AI drives the preview through the steps.