A verifiable AI execution infrastructure that wraps each model inference in a cryptographically sealed Certified Execution Record (CER). One SDK call (npm install @nexart/ai-execution) canonicalises the inputs, outputs, parameters, and context, then hashes them into a certificateHash so any later alteration breaks the hash. Verification happens at verify.nexart.io in the browser, with no account or API key, and can be done fully offline given the CER and public key. An optional NexArt node countersigns records with Ed25519, but verification never depends on trusting NexArt. Multi-step agent workflows can be bundled into a single verifiable artifact identified by a projectHash. The scope is explicitly bounded: it proves a record has not been altered, not that an AI output is correct, fair, or lawful—it is an evidence layer beneath observability, logging, and evaluation tools, not a replacement for them. Operated by Artnames Ltd (UK, registered September 2025), founded by Jeremy Bouedo.
Pros & Cons:
Pros: Cryptographic sealing plus offline verification means a third party (auditor, client, regulator) can validate what the AI did without trusting either the operator or NexArt; the integration is a single SDK call rather than a proxy or infrastructure change, lowering adoption friction; optional Ed25519 node countersigning adds third-party attestation without making verification dependent on the vendor; project bundles give multi-step agent workflows a single verifiable artifact; the scope statement is unusually honest—it explicitly does not claim to judge correctness, fairness, or legality; the target use cases (SOC 2, ISO 42001, EU AI Act evidence) address a real and growing compliance need.
Cons: Very early-stage—Artnames Ltd was registered in September 2025 and the product is still refining its protocol and integration surface with early adopters, with no public customer case studies or independent user reviews; pricing is opaque—no public pricing page surfaced, and the $6,000 starting figure from a third-party directory cannot be cross-verified on the official site; **the G2 listing appears to describe a different company** (digital solutions, software development), creating search confusion; **verification is not correctness**—users expecting AI quality evaluation will be disappointed, and the product depends on developers deciding which steps to record; no public SOC 2 or ISO certification for NexArt itself, despite positioning around compliance evidence.
**Who is it for?**:Teams deploying AI agents in regulated or audited contexts that need tamper-evident execution records; companies preparing for SOC 2, ISO 42001, or EU AI Act evidence requirements; platforms whose customers or auditors demand independent proof of what an AI did; developers building multi-step agent workflows that need a single verifiable artifact per run.