Capture
Enter the model identity, revision, prompt, output, seed, and optional notes. Content remains in this browser.
Local-first AI evidence
Prompt To Proof
Package an AI run into one deterministic receipt. Kernave hashes model context, prompt, output, and run metadata in your browser—ready to export, inspect, and verify again.
An AI output is easy to copy. Its context is easy to lose. Kernave keeps the model, input, result, and run notes together. One canonical body becomes one SHA‑256 digest. Anyone holding the file can check whether it changed.
The receipt system
01 / Why a receipt
A stable record around a probabilistic result.
Kernave does not pretend to prove that an AI answer is true. It proves whether the supplied run record has changed since the digest was created. That narrower claim is useful: teams can move an intact envelope from a model runner to review, audit, publishing, or a later network anchor.
Canonicalize
Keys are ordered consistently and text is normalized before hashing, so the same body produces the same digest.
Verify
Recompute the digest locally at any time. A changed field produces a mismatch instead of a cosmetic success state.
Extend
Export the open receipt format, sign it with a wallet only when requested, or prepare a guarded anchor envelope.
02 / Working product
Build a reviewable AI run.
Use real run details or load the clearly marked sample. Creating, exporting, importing, and verifying a receipt happens locally. No prompt or output is sent to an API by this page.
Your generated receipt will appear here. A digest proves integrity, not the truth or quality of the AI output.
03 / Web3 layer
Keep identity, points, payment, signing, and anchoring separate.
Wallet connection only reads the selected address. Points activation is a second action. The optional 0.42 BOT upgrade has its own preview. Signing and anchoring are separate again, so one click never hides multiple permissions.
Built for an open Web3 ecosystem