API

Validate a prompt

POST /api/validate scores an arbitrary system prompt against the same rubric as /validate and the MCP server's validate_prompt tool — all three go through one guard chain and share one rate limit, so a status you get from curl matches what the page would show.

bash

1curl -X POST https://app.rubricary.com/api/validate \
2  -H "Content-Type: application/json" \
3  -d '{"text": "You are a support agent for Acme..."}'

A successful call returns the score, the criteria it was checked against, and how many free runs remain:

json

1{
2  "score": 78,
3  "band": "good",
4  "findings": [ /* one entry per rubric criterion */ ],
5  "freeRemaining": 1
6}

No key is required and none is accepted — the endpoint is anonymous by design, metered by IP and account instead:

Status	Reason	Meaning
400	—	Request body is missing or empty `text`.
401	sign_in_required	The 2 free anonymous validations are used up.
413	too_long	Prompt text is over 12,000 characters.
429	ip_quota / user_quota	Daily quota reached — 10 per IP, 50 per signed-in account. Resets at midnight UTC.
503	budget_exhausted	The validator's model budget for the period is spent.
502	—	The configured provider could not be reached.

Full detail on what happens to the text you send — what's kept, what's hashed, what's never stored — is in What the validator does with your text.