Skip to main content
No. Tests are written as sentences. The API and CLI exist for people who want to run them from a pipeline, and are entirely optional.
No. Testorim drives a real browser against your public address, exactly like a visitor. There is no script to embed and no agent to run.
Not directly. Private and internal addresses are refused, deliberately. Expose the build through a tunnel and use the public URL.
Yes. Every plan starts with a 3-day trial that needs a card: 10 runs, 30 browser minutes, one run at a time, and nothing charged until the trial ends. See Plans and limits.
Fewer than you might think. A saved procedure replays without planning, so the steps are fixed. Only a chat-driven run is planned; a replay still makes a small number of calls for the report and for resolving a target that is not exactly named. Those calls count against the plan’s AI execution allowance, which is why the advice is to save the tests you run repeatedly.
Chromium, through Playwright, headless. On the Team plan a saved procedure can also replay in Firefox and WebKit (the engine behind Safari) when you run it from the app.
Yes, up to your plan’s concurrent-run limit (1, 2 or 5). Over it, a run is refused at once rather than queued. Separately, when every test browser in the service is busy, a run waits in line and the chat shows its place.
In a run you are watching in the app, the run stops and asks you, and continues with what you type, up to a 10-minute wait. Scheduled, API, CLI and PR runs have nobody to answer, so do not use it there. For those, a login-setup procedure is the way.
Not in the clear. Values typed into password-like fields are redacted before the run is persisted, before the report model sees them, and before a share link renders. When you save a run as a procedure, the password it typed is stored encrypted so the replay can type it, is shown as *** everywhere, and is typed only on the project’s own addresses. Environment headers, database connections, browser cookies and saved login state are encrypted at rest.
Yes. A share link is a revocable, optionally expiring public URL to one run’s report and recording, with credentials, environment headers, browser state and console/network detail stripped.
Because the action worked but something is worth knowing: the app moved the page mid-step, an Enter key was ignored and a click stood in for it, a click changed nothing, a target had to be inferred. Warnings do not change the verdict; they are what makes a green run trustworthy.
Because it was not. A check you asked for that no step covers becomes a visible skipped step, and the run is marked Needs review instead of passing. Usually the test is doing too much, so split it.
Every step that ran passed, but the run could not confirm that what you asked for was met: a check never ran, a check proved nothing, or the report found the result did not match. It is neither a pass nor a failure; a person has to look. See Run status.
Yes. Export as code on a procedure gives a standalone Playwright spec. Passwords and variables are read from environment variables, never written into the file.
No. Trigger runs yourself with the API or CLI and read the result. The CLI’s exit code is 0 on pass, 1 on fail, 2 on error, 3 when the run needs review.
New runs are refused once the subscription ends. Projects, procedures and settings stay. Run history then follows the retention of an account without a plan, 7 days, so older runs are deleted.
According to your plan’s history retention: 30, 90 or 365 days. See Plans and limits.