Smithery is now a part of Arcade.dev! Read more in our announcement here
Execute Unity Test Runner. Use when: running unit tests, verifying code changes, checking test results, or when user asks to run tests. Returns test results with pass/fail counts.
Execute Unity Test Runner. When a test or suite fails or a test ends inconclusive, NUnit XML results with failure messages, stack traces, and inconclusive reasons are automatically saved. Read the XML file at XmlPath for detailed diagnosis.
uloop run-tests automatically compiles pending script changes before running tests. Pass --skip-compile only while validating active hot-reload patches, because the compile clears those patches; otherwise let the default compile surface errors and run against current scripts. --skip-compile skips only the CLI-side compile: script edits saved since the last compile can still be imported during the run; such an import reloads the domain as soon as the run releases its assembly lock, discarding active patches and ending the request. While hot reload holds Auto Refresh (the state an apply leaves), a verified run kept the patches live. Check uloop hot-reload --status after the run instead of assuming either outcome.
Before executing tests, uloop run-tests handles unsaved loaded Scene and Prefab Stage changes according to --unsaved-changes (default save): save writes them first, fail stops if any remain, and discard reloads disk state so tests run against saved files. Untitled scenes cannot be discarded and fail. If the chosen mode cannot proceed, it returns Success: false, keeps TestCount at 0, lists the items in Message, and does not start the Unity Test Runner.
Active pause points are automatically cleared (the underlying code patches are removed as well) before test execution begins. Cleared IDs are reported in the response's ClearedPausePointIds field.
A test run can end by discarding active hot-reload changes: script edits imported during the run are compiled when the test runner releases its assembly-reload lock, and that deferred domain reload wipes the patches even though the tests themselves ran patched. The response's Warning field appears whenever changes were live when the run started and does not say whether they survived; check 'uloop hot-reload --status', then re-apply 'uloop hot-reload' or bake the edits in with 'uloop compile' before the next Play or test run.
NoTestsFound means zero tests matched — not a test failure. Check NoTestsFoundExplanation and Message for asmdef hints. When an unfiltered run finds no tests and the project has no test assembly for the TestMode, ProposedTestAsmdef carries a ready-to-write .asmdef: save Content at AssetPath, move the test scripts under that folder, then compile and rerun.
Status: Inconclusive means no test failed but at least one could not meet an Assume; Success is false, as Unity's batchmode test run also fails for it. InconclusiveTests names them with the assumption's message. A test that cannot run in this environment should call Assert.Ignore so it reports as skipped.
uloop run-tests [options]
| Parameter | Type | Default | Description |
|---|---|---|---|
--test-mode |
string | EditMode |
Test mode: EditMode, PlayMode |
--filter-type |
string | all |
Filter type: all, exact, regex, assembly, class |
--filter-value |
string | - | Filter value (test name, pattern, assembly, or class name) |
--unsaved-changes |
string | save |
save writes unsaved Scene/Prefab Stage changes; fail stops if any remain; discard reloads disk state (Untitled scenes fail) |
--skip-compile |
flag | - | Skip the CLI-side compile before running tests; use only while validating active hot-reload patches. Script edits saved since the last compile can still be imported during the run. |
--timeout-seconds |
integer | 600 |
Maximum seconds to wait for RunFinished before canceling the await (max 1500). Increase for long suites; on timeout the Test Runner may still be running until stop handling lands |
--respect-enter-play-mode-settings |
flag | - | PlayMode only: keep the project's Enter Play Mode settings instead of forcing Domain Reload off. A Domain Reload during the run is survived; the result is recovered after the reload. Use for projects whose libraries require a Domain Reload on Play entry. |
--rerun-failed |
flag | - | Rerun only the tests that failed or were inconclusive in the most recent completed run of the same --test-mode (whole fixtures for a failed OneTimeSetUp/OneTimeTearDown). Cannot be combined with --filter-type or --filter-value |
By default PlayMode still forces Domain Reload off. With --respect-enter-play-mode-settings, a Domain Reload may run and the command takes longer; pause-point and hot-reload notes are omitted from a result recovered after reload. Canceling the CLI (Ctrl-C) does not stop the Unity-side run on this path.
exact matches the full test name (Namespace.Class.Method). class runs every test of one class by bare or namespace-qualified name, e.g. --filter-type class --filter-value PlayerTests; the name is matched literally and whole, so PlayerTests does not run EnemyPlayerTests. regex matches a .NET regex against full test names, e.g. --filter-type regex --filter-value '^MyGame.Tests.'
--rerun-failed reads the record that every completed run writes for its test mode, so it reruns the failures of the most recent completed run, filtered or not. A run that timed out or was cancelled leaves no record. With nothing recorded as failed it returns Status: NothingToRerun without running; to tell a flaky test from a real failure, run --rerun-failed again after a failure.
Returns JSON. Success, Status, Message, FailedTests (up to 10), and XmlPath (set when a test or suite failed or was inconclusive) are usually enough; every field is described in references/response-fields.md, and the XML file in references/xml-results.md.
