NEAR Hackathon 2026 · Privacy & Private AI · Agent Economy
Prove an agent is good for it, without handing over its wallet
Privacy is the point. A wallet address is not just a number. Anyone who sees it sees every balance, every transfer, every counterparty, and publishing a treasury wallet to prove you can pay invites targeting, scams and front running. VeilProof proves an agent can pay without ever revealing its wallet. The public sees only hashes and claims, never the address, never an exact balance. Agents are starting to pay each other: compute, invoices, OTC deals, and before a deal, the other side wants proof the agent can pay. VeilProof proves the treasury across every wallet the agent holds, proves the balance was HELD and not just deposited an hour ago, and carries a portable credit score any app can recompute from the same public rules. The exact figures go only to the verifier the agent chose, in a private package, and a fresh random salt per report means two reports from the same treasury cannot be linked. The facts come from the chain itself, so they cannot be faked.
Report loading…
| Aggregate balance at least 10 NEAR | Pass |
| Sustained balance at least 10 NEAR | Pass |
| Volume in 30 days | Pass |
| Transactions in 30 days | Pass |
| Credit score | … |
Underwriting (advisory): loading…
Anchored on NEAR testnet · run the live demo
What stays private
Privacy is the point, not a feature. Here is exactly what the public sees, and what it never sees.
The address never goes public
Only a hash commitment of the wallet is ever published. The address itself travels only in a private disclosure package to the verifier you chose, and only because you shared it with them.
Exact values stay sealed
The public report carries thresholds and a score band only, never an exact balance. The exact figures are sealed for the verifier alone, and a tampered report fails verification.
Reports cannot be linked
A fresh random salt per report means two reports from the same treasury share no identifier. The public cannot tell they came from the same wallet.
Measured, not promised
We ran a privacy audit with 50 simulated provers. From the public report alone, an attacker recovered 0 of 50 addresses, and its balance guesses were off by 94 percent. The numbers are in the README, not a slogan.
One honest limit: the privacy is from the public, not from the verifier you chose. They must see the address to recheck the chain, so you share it with them and no one else. Read the full explanation.
Why VeilProof
Proof a counterparty can check, without the exposure that usually comes with it.
Private by design
Only hashes go public. The wallet address travels only to the verifier you chose, in a private package, and a fresh salt per report keeps reports unlinkable.
Genuine by construction
The facts come from the chain itself, so they cannot be typed in or faked. The verifier rechecks every claim from the same data the report was built on.
Portable credit score
A deterministic score any app can recompute from the same public rules. No black box, no trust needed, and a tampered score fails verification.
How verification works
Three steps. The prover never reveals the wallet to the public; the verifier never has to trust the prover. Verification answers five separate questions, and each one gets its own status.
Read the full explanation, including what the verifier can and cannot learn.
What it proves, and what it does not
The honest trade off, stated plainly. VeilProof is a hybrid privacy model: the public sees only hashes, the chosen verifier sees the address because they must recheck the chain, and the AI sees claim aggregates only, never the address, never transaction hashes.
| Claim | What proves it |
|---|---|
| The data is genuine | The verifier checks the disclosed snapshot against the data hash published onchain. Onchain data cannot be typed in by the user. |
| The balance was held, not flashed | Sampled balance history at past block heights: the lowest sampled balance over the window, not a snapshot. |
| The wallets are all the prover's | Each linked wallet proves ownership with a transfer to the primary wallet, or by signing a link message with its key, no transfer, no gas. |
| The credit score is real | The score is deterministic (score v2, published rules): the verifier recomputes it from the same snapshots and the same windows. A tampered score fails verification. |
| The underwriting opinion is honest | NEAR AI Cloud TEE attestation is checked server side when the analysis runs: the DCAP quote, the nonce, the response signature and the measurements. This page does not recheck them, so the attestation status is shown on the verifier page. The verdict is advisory, a thin score can never support a lend, by a deterministic guard. |
| The report was not altered | SHA 256 report hash anchored on NEAR; tampering fails verification. |
| The wallet belongs to the prover | The prover signs a challenge with the wallet key; the publish transaction records the signer as owner. |
| The wallet address stays private | Only a hash commitment is published. The address travels only in the private disclosure package to the verifier the prover chose. |
| Anything offchain | Nothing. VeilProof only ever makes claims about onchain data. |
Who uses it
Ready to prove it?
Preview any wallet with no wallet connected, or run the live demo and watch a real claim verify against the testnet registry.