The NexArt Protocol for Verifiable AI Execution

The NexArt Protocol is an open execution and verification standard for deterministic and integrity-bound systems. It defines how executions are captured, sealed with cryptographic hashes, and independently verified, across two distinct execution surfaces under one versioned trust model.

Two Execution Surfaces

Code Mode provides deterministic, reproducible execution where the same inputs always produce the same output. AI Execution produces tamper-evident records of AI and LLM calls: integrity evidence that anyone can verify independently. Both surfaces share the same canonical hashing rules and attestation model.

Why Verifiable Execution Matters

Logs can describe what happened. They cannot prove it. Execution drift creates silent risk in AI agent decision workflows, compliance-sensitive pipelines, financial simulations, and model evaluation systems. Verifiable execution closes the gap between operator claims and independently checkable evidence.

Protocol-Level Guarantees

How Verification Works

  1. Capture: inputs, parameters, outputs, and context recorded at runtime
  2. Seal: fields canonicalized and bound with a SHA-256 hash
  3. Attest: independent node issues an Ed25519-signed receipt
  4. Verify: anyone recomputes the hash and confirms independently

Protocol Architecture

The protocol separates the canonical specification (what must be true for a record to verify) from reference implementations (how to produce conforming records). This keeps results portable across SDK versions, runtimes, and attestation nodes that conform to the same protocol version.

Applicable Domains

Protocol Status

The current protocol version is v1.2.0 for Code Mode and ai.execution.v1 for AI Execution snapshots. Versioned guarantees are documented in the public specification at docs.nexart.io.

Certified Execution Records · AI Execution Integrity · Get started with the SDK