Which way should I use ClearRun: npm kit, API, or chat?
Use the npm kit in mock mode while you build. Use the API when you already have answers from any model and want them scored. Use chat when you want ClearRun to write the answer and score it in one step.
Last updated
Compare the three ways
| npm kit (SDK mock mode) | Hosted API | Chat | |
|---|---|---|---|
| What it does | Returns local test scores on your computer | Scores text you send | Writes an answer, then scores it |
| Needs an account | No | Yes, with an API key. A verified free account can create one | Yes. Any plan on the site. API chat can use that key |
| Uses credits | No | Yes | Yes |
| Real scores | No | Yes | Yes |
| Signed, with a receipt | Never | Yes, for each successful call | Yes, for each successful call |
Pick by what you are doing
- Building an app or tests: start with the SDK in mock mode. Add a key when you need real scores. See How to use ClearRun mock mode for local development.
- Scoring answers you already have: call
POST /api/evaluatewith the answer and the prompt, orPOST /api/scorewith the answer alone. The CLI, SDK and MCP server all call these for you. - On a free account: sign in and paste text into Paste Check, or use chat on the site. Neither needs a key.
- One step, question in, scored answer out: use chat on the site, or call
POST /api/chat. See How to get started with ClearRun chat.
What none of them do
None of them check facts. The score is about whether an answer's confidence is backed by its reasoning. It is not a check that the answer is true.