POST /v1/research
A question in, a verified answer out: sources, the passage Jev read, its character span, and a support score per paragraph.
Put a fact-checker in front of your model.
The Zearch API gives your product verified research and paragraph-level claim checks, with the exact evidence attached.
POST /v1/verify
Your model's text and your sources in. Each paragraph comes back supported or not, so you can block it before a user sees it.
Built to retry
Idempotency keys, streaming, a single error shape and rate-limit headers. Same credits as the app.
Ask
curl https://zearch.computer/v1/research \
-H "Authorization: Bearer $ZEARCH_KEY" \
-H "Content-Type: application/json" \
-H "Idempotency-Key: $(uuidgen)" \
-d '{"query": "When did the James Webb Space Telescope launch and where does it orbit?"}'Verify
curl https://zearch.computer/v1/verify \
-H "Authorization: Bearer $ZEARCH_KEY" -H "Content-Type: application/json" \
-d '{"text": "At sea level, water boils at 100 °C. [1]",
"sources": [{"title": "Boiling point", "text": "Pure water boils at 100 °C at one atmosphere."}]}'
{"object": "verification", "passed": true, "threshold": 0.75,
"paragraphs": [{"index": 0, "support": 0.97, "attribution": 0.93, "passed": true}]}Pricing
API calls spend the same credits as the app: Search 1, Compare 2, Deep 5, Verify 1. Errors are free. See plans.