Anchor an agent output

Paste the output of an AI agent, model run, or generated draft. It is fingerprinted in your browser — the text itself never leaves this page. Only the SHA-256 is submitted and committed to Bitcoin via OpenTimestamps; an optional SHA-512 sibling witness is recorded alongside it in the receipt. The receipt is later verifiable by anyone, offline, without Orphograph.

0 characters · hashed locally with WebCrypto

The manifest itself stays with you; the receipt records only its hash, so the two provenance trails corroborate each other.

What this proves — and what it doesn't

The receipt is cryptographic proof that this exact output existed before a point in time. It is provenance, not detection: it makes no claim about whether content is human- or machine-made, and no claim of authorship. Agents can anchor programmatically through the MCP server — this page is the same anchor, by hand.

A receipt created through the API or the MCP server can additionally carry a zero-knowledge execution proof — evidence that a fixed, published hash-chain procedure produced the output while the inputs stay private. Its scope is stated plainly: it proves a fixed hash-chain procedure ran — not that a specific AI model ran. And the proving key ceremony is development-grade; we do not yet make this claim in any certified sense.