Create with AI. Or record yourself.
Any AI that can write plain text can help create a .testhand test. Prefer doing it yourself? Click, type and pick menus in the Qt app while the recorder writes your supported actions into the same readable format.
FROM A MOMENT TO A METHOD
Do it once.
Make it repeatable.
Create a test with AI, or record the workflow yourself in the Qt app. Add checks for what the app should show—not just the actions you perform—then replay the test when the application changes.
Add a check. Not just an action.
While recording in the Qt app, press Check and click the result you want to verify. TestHand records its current value, approved text or picture—not another click on the control. Hold Shift to check its picture.
Review the steps and results.
AI-written and recorded scripts stay editable. Verify selectors against your app and review expected results. An AI connected through MCP can inspect and operate a correctly configured local app; a plain-text draft still needs a real replay.
Replay. See what happened.
Stop recording, press Stop, then Play from the start. A failed step turns red and holds. Why it failed shows the error and failure screenshot; Play retries the failed step.
A SMALL FIRST WORKFLOW
Say hello.
Then test the result.
Launch and connect to a Qt app with these example controls before running the script. Replace selectors with your application’s actual names.
smoke.testhand / plain text
# Launch and connect to your app first.
type 'Ada' --target '#nameEdit'
click '#greetButton'
wait '#greetingLabel' --for property \
--property text --value 'Hello, Ada!'
screenshot --out result.png01 / Perform the action
Type “Ada” and click the greeting button using selectors from the app’s object tree.
02 / Wait for the outcome
Wait for the greeting property instead of adding an arbitrary pause.
03 / Keep the evidence
Capture the window. Open the sequence in the visual player to inspect steps and failures.
Try the browser replayRECORD IN THE PLAYER
From your hand
to a readable file.
Choose File → New Recording, select your Qt application and save a new .testhand script. An optional description explains what it tests.
Use the app normally.
Click, type, choose menus, scroll and drag. Supported actions appear as steps with a red dot. For an existing script, Record replays to the end and appends new actions before any final quit steps.
Keep the app visible.
Recording is unavailable while the app runs hidden or a group plays. Hover-only actions and native macOS menus are not recorded. Do not assume the external desktop backend has the Qt recorder’s capabilities.
CHECK WHAT THE APP SHOWS
One click.
A repeatable check.
While recording, press the tick-shaped Check button next to Record, or choose Play → Check (Ctrl+Shift+K). The pointer becomes a crosshair over the app.
Select the result, not another action.
Click the control you want to check. The selection click does not reach the app. TestHand adds a step that verifies what the control shows now each time the script plays.
Choose a picture when it matters.
Hold Shift while clicking to check the control’s picture instead of its text. Press Esc or Check again to cancel. Password fields are never checked.
Values
For a label, text field, check box, drop-down list, spin box or slider, the recorded step waits for the current value—for example, “42 photos” or a ticked check box.
Approved text
For a list, table, tree or whole panel with a text snapshot, the check compares its text with the approved result saved next to the script. When no text snapshot is available, it uses a picture.
Pictures
For a picture, drawing area, other controls or a Shift-click, the check compares against the picture saved when you recorded it. Keep the script’s .approved folder with the test so its reference text and pictures remain available.
PROVE THE RECORDING
Replay from the start.
Inspect what fails.
Press Record again to finish recording, then Stop and Play from the start. Newly recorded steps already count as done, so continuing playback alone does not test them again.
On screen, or out of your way.
Leave Show the app on to watch the test run on screen. Turn it off for supported offscreen playback when the app’s windows would be distracting. The test can run without showing the app, leaving your screen free for other work.
Offscreen support depends on the application and backend. External desktop automation can still use real windows, keyboard and pointer input; do not treat it as hidden playback. Turn Show the app back on before recording.
Go at your pace.
Continuous mode advances automatically. Step by step waits for Play after each step; Step runs one step. Add a breakpoint to hold before a step.
Investigate, then retry.
A failed step turns red and holds. Open Why it failed for the error and screenshot captured at failure. Play retries that step; double-clicking another step continues from there. Review the expected result before changing the test.
FROM AN IDEA TO A REPEATABLE TEST
Your next test.
A little more human.
Choose the right backend, connect to your application and start with one small workflow.
Get started