Any escrow system eventually has to answer the same hard question: when the buyer and seller disagree, who decides who's right — without reintroducing the centralized company you were trying to avoid in the first place? Kleros and VaultPay both answer that question with the same basic idea — let a group of independent people vote, and let the contract carry out whatever they decide — but the mechanics diverge in ways that are worth understanding if you're evaluating either one.
Kleros Escrow: a general-purpose court, secured by a token
Kleros isn't specific to escrow — it's a broader "decentralized court" protocol that plugs into many different kinds of dApps needing an arbitration layer, escrow being one use case among several. Its jury system works like this:
- Anyone can become a juror by staking PNK, Kleros' native token, into a specific court (Kleros runs specialized sub-courts for different dispute categories).
- When a case opens, jurors are selected randomly, weighted by how much PNK they've staked — stake more, get selected more often.
- An initial ruling can come from a jury as small as three people. If a party disagrees, they can appeal, which pulls in a larger jury for the next round.
- Votes are submitted via commit-and-reveal (jurors commit a hidden vote, then reveal it later) to prevent jurors from copying each other.
- The incentive to vote honestly is economic: jurors who vote with the eventual majority earn a reward: jurors who don't, lose staked PNK. This is the "Schelling point" mechanism — voting truthfully is each juror's best individual strategy because it's what they expect everyone else to converge on too.
It's a genuinely clever design, and the appeals ladder means high-value or contentious disputes get progressively more scrutiny rather than a single small jury's word being final.
VaultPay's reviewer pool: simpler, smaller, no separate token
VaultPay's dispute mechanism is deliberately less elaborate. There's a single pool of registered reviewers — capped at 10 at a time — and when a dispute opens, 5 of them are pseudo-randomly selected to vote. Each reviewer picks one of three outcomes: 100% to the seller, 100% to the buyer, or a 50/50 split. Whichever option gets a plurality wins (an exact tie between the two extremes resolves to the 50/50 split). No appeals ladder, no separate token to acquire or stake, no commit-reveal step — reviewers vote once, the contract tallies it, and the deal resolves.
Reviewers earn a share of protocol fees for participating, rather than staking a token to win voting rights.
The actual tradeoffs
Sybil resistance and stake in the outcome. Kleros' PNK staking means a juror has real economic skin in the game on every single vote, and losing money for an incoherent vote is a strong honesty incentive. VaultPay's reviewer pool has no equivalent financial stake mechanism — reviewers are simply registered wallet addresses, filtered by a minimum time-registered requirement before a given deal was created (so you can't join the pool the moment a dispute you're involved in opens). That's a real, meaningful difference in the strength of the honesty guarantee, and it's the tradeoff we'd point to first if someone asked us to be critical of our own design.
Pool size and decentralization at small scale. A 10-person cap is small, especially compared to Kleros' much larger, cross-application juror base. With few concurrent disputes this is manageable; it's a real constraint if usage grows and many disputes are open at once, since the same small set of people ends up adjudicating most of them. This is a known limitation we're actively thinking about, not something we're glossing over.
Simplicity and cost. There's no token to buy or stake to become a VaultPay reviewer, no gas cost for a commit-and-reveal round trip, and no appeals process to navigate. For a platform in early beta, that's a deliberate simplicity tradeoff — fewer moving parts, lower friction to participate as a reviewer — made at the cost of the stronger guarantees Kleros' more elaborate system provides.
Scope. Kleros arbitrates disputes across many different dApps and use cases through its general court system. VaultPay's dispute mechanism only exists for VaultPay's own escrow deals — it's purpose-built, not a general arbitration layer other projects can plug into.
Which one is "better"?
Neither, in the abstract — they're optimized for different things. Kleros trades complexity and a token requirement for stronger economic guarantees and a system that scales across many applications. VaultPay trades some of that rigor for a simpler, lower-friction reviewer experience purpose-built for one thing: escrow disputes between a buyer and a seller. If you're building something that needs general-purpose, high-stakes arbitration, Kleros' model is more battle-tested. If you're a buyer or seller on VaultPay wanting to understand who's deciding your dispute and why, this is the honest shape of it — including the parts we'd improve given more scale.
VaultPay is currently live on Base Sepolia (testnet). See how a VaultPay deal and dispute actually works end-to-end.