NiyamAI - An Intent-Bound AI Agent with Cryptographically Verifiable Guardrails using Zero-Knowledge Proofs

NiyamAI uses zk-SNARK proofs via EZKL to provide cryptographically verifiable guardrails for autonomous agents. The framework commits permitted tool constraints to a SHA-256 Intent Contract and requires an isolated Judge model to generate proofs for every intercepted tool call before execution. This allows third parties to verify policy enforcement without accessing model weights.