CALIBRATEDAGENTS

Playground

Paste a question, the context it was answered from, and the answer. Every span of the answer is scored and flagged.

Each sentence-level claim in the answer becomes a span. Lookup claims are scored; claims that need arithmetic or comparison are marked unsupported in Basic.
ANSWER · spans coloured by verdict
PER-SPAN
Basic tier. Structural verification of lookup claims — value, scope, polarity, refusal, contradiction. Pro (reasoning: totals, comparisons, branch conditions) is coming.
Each line is treated as a separate passage.
0.0 · answer absentanswer present · 1.0
probability the answer is supported by the context
API access

Want to test it through the API?

Leave your email and we'll send API access details as we open the tier up.