CycleMantle
FAQ

Free api for testing, fixtures and the CycleMantle app

05

Testing basics

What is a free api for testing?

A free api for testing is a public web service that returns sample data, and people who search for it usually want a sandbox to practise against. CycleMantle is not one. It writes fixed replies into a JSON file that you own, so your tests never depend on someone else's server or on whether it is online today. Use a public service only if its owner allows it.

What is a dummy api?

A dummy api is a stand-in service that returns fixed, made-up data instead of live output. In a test project it is often just a JSON file. CycleMantle can write that file for you, with invented names, dates and amounts, and show it as a diff before anything is saved to your folder. Keep it small. Add rows only when a test needs them.

How do I get realistic test data for frontend work?

The simplest way is to ask the agent for a JSON file with invented rows that match the fields your page shows. Describe the columns, the number of rows and a few odd cases such as an empty name. Read the diff. Edit any value you dislike, and keep the file in your project so the next test can reuse it.

What should a coverage report tell me?

A coverage report is a list of the lines or branches that ran while your tests ran, so it points to code that no test touched. Numbers mislead. It cannot say whether the checks were meaningful, so read it as a to-do list and not as a grade, and open the functions with untested lines first. Ask the agent for one test at a time.

How do I test code that depends on a late backend?

The simplest way is a fixture file that holds fixed replies for each call. If you must mock endpoints when the backend is late, ask the agent to write that file and a test that reads it. Keep it honest. Later, compare the fixture with the real replies of a service you own, and update it whenever they change.

Using CycleMantle

Is phptest.org the official CycleMantle website?

Yes. This is the official CycleMantle website, and the download button gives the original build of the app, whose SHA-256 checksum you should compare with the published manifest before you run it. Questions and bug reports go through the Send feedback button inside the app, and the app is independent of the PHP project. You can also read the version number on the download page, which shows 2.9.30.

Are the results always right?

No. The agent can misread code, miss an edge case or write a weak test, so you review every diff and run your test command yourself when in doubt. Passing tests and a high coverage figure are evidence, not proof. Ask first keeps each change waiting for your OK until you have read it. Keep your version control history as a safety net.

Can CycleMantle run my test command?

Yes. Shell runs the command you give it in the environment you chose, which can be local Windows, WSL or SSH to a server you own. The card shows the real exit code, and the full output appears when you expand it. Dangerous commands are blocked, and you can switch to Plan only at any time. Nothing else runs unless you ask for it.

How do I check a function whose output changes?

The simplest way is to give the function a known input, so that its output stops changing. To unit test a function with variable results, such as one that returns the current date or a random number, ask the agent to pass that value in as a parameter and write the test around it. Review the diff. Then run your test command again.

What is the difference between Ask first and Plan only?

Ask first is the mode that pauses before actions that change something, such as writing a file, and it shows the diff to accept. Plan only is the mode that writes a plan and runs nothing, so no command runs and no file changes. Plans survive restarts. Switch the mode when you want the agent to carry the plan out.

Models, privacy and download

Do my files leave the PC?

Yes, with a cloud model: your prompt and the file text the agent reads go to that provider. With a local model nothing leaves your PC. Chats, working folders and API keys stay on your PC, and the app has no telemetry. You choose the endpoint. That choice decides where the text of your code goes. Secrets in tool output are redacted before they reach the model.

Which models can I connect?

A model in CycleMantle is the AI behind the chat, and you can connect 3 kinds: a cloud AI provider through a standard API endpoint, a model running locally on your PC, or the built-in model, which is ready to use without your own endpoint. Press the Test button after entering an endpoint to see the response time and the context window.

How do I download and check the installer?

The simplest way is to press the download button, which gives the setup program of about 34 MB, and then compare its SHA-256 checksum with the published manifest. The download page shows a Windows command for that. The installer is not code-signed yet. Windows SmartScreen may warn you, and you continue only after the checksum matches. Keep the manifest page open while you compare.

Do I need an account to use the app?

No. CycleMantle runs without an account, and its interface comes in 13 languages. A cloud AI provider may need its own account and may charge for usage, but that is between you and the provider. The built-in model is ready to use without an endpoint of your own, so a first test needs no sign-up at all.

CycleMantle for Windows

Try one small test with CycleMantle

Download CycleMantle ↗