Skip to main content
Anchoring runs on Base Sepolia, a testnet, today. Base mainnet anchoring is not switched on yet.
An anchor is a public, timestamped commitment to a log root. Its job is to make the log’s history provable to someone who doesn’t trust QED Proof’s servers at all: once a root is anchored, receipts under it can’t be backdated, quietly edited, or deleted without the tampering being detectable.

​
How it works

QED Proof anchors log roots as attestations through the Ethereum Attestation Service (EAS) on Base. EAS is a predeployed public contract with its own schema registry and explorer — QED Proof writes no custom contract for this. The attestation’s data decodes, under QED Proof’s published schema, to the log’s identifier, the tree size, and the root hash at that size. A receipt’s proof.anchor looks like:
chain is a CAIP-2 chain identifier — eip155:84532 for Base Sepolia, eip155:8453 for Base mainnet.

​
Checking an anchor

Given a receipt’s proof.anchor and the issuer’s /.well-known/poaw-keys.json (which publishes the attester addresses and schema IDs it considers valid), you can look up the EAS attestation directly on a public Base RPC and confirm:
  • the attestation came from an address the issuer actually publishes as one of its anchor attesters,
  • it was made under a schema the issuer publishes,
  • and it decodes to the log_id, tree_size, and root_hash the receipt claims (or, if the attested tree is larger, that a valid consistency proof connects the two sizes).
The attestation’s block timestamp becomes the receipt’s proven-by time — the point after which the receipt could not have been backdated. See check a receipt for running this check yourself with the reference checker.

​
What an anchor gives you, and what it doesn’t

An anchor gets a receipt to trust level 2 (see trust levels): a checker can confirm the receipt existed no later than the anchor time, and that it can’t be backdated or silently deleted. It does not, on its own, confirm what a hardware-attested or web-proven verifier build would (trust levels 3 and 4) — those are specified but not built.