For AI agents and AI applications

Sign every AI action and output cryptographically.

When an AI agent takes an action, when a model generates output that may face audit or legal challenge, sign it at the moment it happens. Multi-model neutral. Three-layer cryptographic timestamps. Built for EU AI Act Article 50 disclosure and FRE 902 admissibility.

AI Provenance details Read the docs

When AI accountability matters

Regulated industries

Healthcare, finance, government, any AI output that may face audit or litigation needs a verifiable trail showing what the model produced and when. Self-attestation by the model provider isn't enough.

EU AI Act Article 50

Article 50 enforces August 2026, providers of generative AI must mark output as AI-generated in machine-readable form. Cryptographic signatures are the strongest available form.

Multi-model stacks

If your product calls Claude + GPT + Mistral + open-source models, you need a neutral third-party signer that doesn't favor any one provider. AI Provenance is multi-model from day one.

Enterprise procurement asking "how do we audit AI?"

Buyers increasingly require AI provenance language in vendor agreements. Having signed receipts on every AI output answers the question before procurement asks it.

How signing works

One SDK call. One MCP tool. One REST endpoint. Pick whichever fits your stack.

// AI agent takes an action

const response = await claude.messages.create(...)

const signed = await cert.signAIOutput({

output: response.content[0].text,

model: 'claude-opus-4-7',

provider: 'anthropic',

})

// signed.receiptId verifies at certnode.io/verify/provenance/[id]

Sign your first AI output in under 5 minutes.

100 signings/month free. Then $0.01/signing. Volume discounts auto-apply.

AI Provenance details