Skip to content
Veral · Verification Authority Layer
Read · Compute · Attest

How Veral works.

Veral reads every public source the engine can reach about a subject, computes a deterministic score on a published formula, and signs the answer on Ethereum. The worked example below is an AI-Agent reading — Veral's flagship cert.

01 · Read

Every public source the engine can reach.

For each subject Veral fetches the public evidence in parallel — source-verification registries, on-chain history, ENS records, EAS attestations, audit registries, identity stamps, social graph, security signals. Every fetch is cached with a published TTL and carries provenance so any reader can re-fetch the same evidence and reach the same conclusion.

Worked example · freysa — an autonomous on-chain agent

On the worked example below, the engine reaches usable evidence across most of its wired AI-Agent-profile sources. Each line is one extractor and the finding it produced.

  • Safe multisig · 2-of-3 Safe controlling the agent address
  • Upgradeability · non-proxy contract, no admin-key escape hatch
  • EAS attestations · three attestations on file, two from trusted issuers
  • ENS · multi-month-old ENS name with two published subnames
  • ENS records · five populated text records — discovery, links, descriptors
  • Ethereum mainnet · live contract deployed, sustained transfer activity
  • L2 deployments · verified deployment matched on one L2
  • Counterparty graph · 14 unique counterparties over 90d, no wash pattern, no sanctioned address
  • GitHub · public repo with tests, CI, license, releases shipped in the last year
  • Gitcoin Passport · self-asserted passport with 12 stamps
  • Farcaster · Farcaster ID with two verified address bindings
  • POAP · 8 unique event POAPs, most recent earlier this year
  • Sourcify · two contracts verified end-to-end
  • Etherscan · every controlled contract source-verified
02 · Compute

A deterministic formula, no AI, no opinions.

The aggregator combines the evidence into a per-macro score, weighted by the methodology's locked weight vector, then rolls up to a single 0–100 number. Same evidence in, same score out, on any device, at any time. Conditional macros that have no signal on this subject re-normalize out — no merit penalty for an absence the methodology declares optional.

Worked example · per-macro trace
Macro
Name
Weight
Score
Eff. weight
M1
Ownership & Control
0.22
0.59
0.27
M2
Identity & Verification
0.14
0.34
0.17
M3
Operational Maturity
0.18
0.63
0.22
M4
Financial Standing
0.15
n/a
0.00
M5
Counterparty Quality
0.14
0.56
0.17
M6
Developer Provenance
0.13
0.49
0.16
M7
Transparency
0.04
n/a
0.00
Total · 56 / 100
Band · Limited
Confidence · High
Coverage · wide source coverage, 5 of 5 macros
The Veral score scale

Every certificate reads back as a 0–100 number on the same evidence-bound scale.

02040607590100
  • 0 – 19 · Adverse
  • 20 – 39 · Caution
  • 40 – 59 · Limited
  • 60 – 74 · Established
  • 75 – 89 · Strong
  • 90 – 100 · Exemplary
03 · Attest

Signed and published on Ethereum.

The reading is published as an EAS attestation anchored to the subject's ENS namehash. The on-chain certificate is verifiable by any reader who recomputes the score against the published evidence bundle, and shareable as a verification URL that is fixed for the life of the attestation on the issuing chain — currently Ethereum Sepolia, a testnet.

Worked example · cert object
Schema
EAS · Veral AI-Agent Cert v1
Network
Sepolia (during private beta)
Subject namehash
0x6e6cb1c7…687b95c
Operator
veral.eth (verified against operator key)
Anchored to
subject ENS namehash
Verification URL
/cert/<UID>
04 · Try it

See it live. Get yours.