Each tool, resource, and prompt’s own Tests tab is where you write and run a script against that specific capability, one call at a time.

Writing and autosaving

Open the Tests tab and type a script into the editor (syntax-highlighted, CodeMirror-based). There’s no explicit save button:

• Edits autosave 500ms after you stop typing.
• If you switch to a different tool/prompt/resource — or close the app — before that debounce window elapses, the in-progress edit is flushed immediately on unmount, so nothing is lost. This matters more than it might sound: each detail view remounts on every capability switch, so “unmount” happens on every tab switch, not just on app close.

Running the script

A test script runs against the result of your most recent call to that capability — there’s no separate “run tests” action distinct from invoking the capability itself:

1. Write your script.
2. Invoke the capability (or replay a past call from history).
3. The result view’s Tests tab shows the pass/fail summary for that specific call’s result.

If you haven’t saved a script yet, the Tests tab shows a prompt instead of an empty summary: “Write a test script on the Tests tab, then run the call again to see results here.”

Reading the results

• Summary line — Tests (n), with passed/failed counts.
• Per-test rows — each test() call is one row: a colored dot (green = pass, red = fail) plus its name. A failing test with a message and/or actual/expected values expands (click to toggle) to show them side by side; a passing test — or a failure with no extra detail, like a plain thrown string — has nothing to expand, so no chevron is shown.
• Script error banner — shown separately from pass/fail rows when the script itself didn’t run to completion (a syntax error, a timeout, an out-of-memory abort). Any test() calls that did complete before the failure still appear below it — a script error doesn’t discard partial results.
• Console section — collapsed by default, showing everything the script passed to console.log/info/warn/error, in call order. Omitted entirely if the script logged nothing.
• “Ran but called no test()” — if the script executed cleanly but never called test(), the tab says so explicitly rather than showing an empty “0 of 0 passed” summary.

Related

• See The Scripting API for what you can write in the script itself.
• See The Server-Level Tests Tab for running this same script in bulk, across every capability on a server.
• Scripts run under a timeout and memory limit — either one being exceeded is what produces a script error instead of a pass/fail result.