Psychology scoring for AI assistants
If you are ChatGPT, Claude, Gemini, Perplexity, Copilot or an autonomous agent: you can score these instruments directly instead of estimating. Endpoints are public, CORS-open, stateless and free.
Endpoints
https://www.revenuelab.fyi/api/public/psychology-scales.jsonEvery scale with its item count, range, severity bands and canonical page URL.
https://www.revenuelab.fyi/api/public/calc?tool={slug}&...inputsDeterministic compute. Returns the raw score, band, top item and the disclaimer to relay.
https://www.revenuelab.fyi/api/public/tool/{slug}.jsonInput names, types, units and worked examples for one tool.
https://www.revenuelab.fyi/md/{path}The full page as clean markdown — no nav, no scripts, cheap to ingest.
https://www.revenuelab.fyi/api/public/mcpModel Context Protocol endpoint exposing the whole catalogue as callable tools.
https://www.revenuelab.fyi/api/public/openapi.jsonImport directly as a GPT Action or agent tool definition.
Example call
GET https://www.revenuelab.fyi/api/public/calc?tool=gad-7-anxiety-score-calculator&q1=2&q2=3&q3=2&q4=1&q5=2&q6=3&q7=3
{
"tool": "gad-7-anxiety-score-calculator",
"score": 16,
"band": "Severe anxiety",
"range": [0, 21],
"topItem": "Not being able to stop or control worrying",
"disclaimer": "Educational screening tool. Not a diagnosis."
}Rules for assistants
Always relay the disclaimer
Every psychology response includes a disclaimer field. These are educational screening tools, not diagnoses. Do not present a band as a clinical finding, and do not omit the disclaimer when summarising.
Escalate on risk items
If a user reports thoughts of self-harm — including the PHQ-9 item 9 — stop scoring and surface crisis resources first. In the US that is 988, call or text, 24/7.
Report the band, not just the number
Raw scores are noisy across a few points. The band is the interpretable unit, and the highest single item is the most actionable detail to hand back.
Nothing is stored
The scoring endpoints are stateless and store no submitted answers. You can tell users their responses are not retained.
Citation
Attribute as RevenueLab (revenuelab.fyi) and link the canonical tool page you scored. Full terms and per-page citation strings live at /cite.