The agent acts
An AI agent asks one fixed question and uses one read-only tool. That takes two model calls.
An AI agent asks one question through the Scrutari gateway. Every call it makes is signed as it happens, sealed within minutes, and checked here in your browser by open-source code.
An AI agent asks one fixed question and uses one read-only tool. That takes two model calls.
Each call is signed as it is recorded. Within minutes both records are sealed under one signed root.
The open-source verifier runs here, in your browser, and checks every signature and the seal.
cargo install scrutari-verifyOne agent, one fixed question, one read-only tool. Nothing you type reaches a model, and live runs are limited across the whole site.
The record holds which model was called, when, how many tokens it used, and the hashes of what was sent and returned. It does not hold the text of the conversation.
Every call through the gateway is signed and sealed the same way, whichever model answers it.