Cryptographic evidence for what AI agents decide

PlainReal™ produces one signed, tamper-evident record per consequential AI agent decision. Each record is independently verifiable by anyone, with no PlainReal service, no account, and no network access required.

The point is not that you should trust us. The point is that you should not have to.

---

What a record proves, and what it does not

This distinction is the product, so it is stated first rather than buried.

A valid signature proves the bytes are the bytes that were signed, and that nothing has changed since. It does not prove:

• that the decision was correct. A verified record of a bad decision is still a verified record of a bad decision.
• that the record is complete. A signature covers what is present. It cannot show what a producer chose never to record.
• write-once custody. Only the declared retention class is signed. Confirming an object is genuinely under a storage lock requires access to the storage account.
• when something happened. That is a timestamp token's job, and published sample records carry none.
• human non-repudiation. A record showing that a person approved something proves the system recorded that approval tamper-evidently. It does not prove the person personally signed anything.

Anyone claiming otherwise about this system is overstating it.

Verify it yourself

verify.plainreal.com checks a record's signature entirely in your browser. It makes no network request and stores nothing. Open the developer tools and watch.

The page is open source, so you can read it before you run it, and each release is a signed tag naming the SHA-256 of the exact page bytes published.

RFC 8785 canonicalisation · FIPS 202 SHA3-256 · RFC 8032 Ed25519

A genuine, signature-valid sample record is published. Edit any byte and its signature breaks, which is the point.

Repositories

verify	The public verifier. Verification only, deliberately: it reads one record, canonicalises it, hashes it, and checks a signature. Nothing else.
sp800-53-mapping	NIST SP 800-53 Rev 5 control mapping for AI agent decision evidence artifacts. Which controls a record actually speaks to, stated per control rather than as a claim of coverage.

More

• plainreal.com: what this is for, and who it is not for
• Regulatory coverage: OCC, NYDFS, EU AI Act, GDPR, SOC 2, NIST SP 800-53, FDA, HIPAA
• EU AI Act Article 12: record-keeping requirements and a readiness self-assessment
• Security: OWASP Agentic AI Top 10 mapping
• llms.txt: for agents reading this

There has been no third-party security audit. Patent pending.

_{PlainReal is operated by AgentOS Technologies, Inc., a Delaware corporation.
contact@plainreal.com · security@plainreal.com}