What TRITO sells
A rollup runs its own chain and does not want another. What it may lack is a set of independent parties willing to attest to something, chosen by a method nobody controls. That is what TRITO rents out: not consensus, and not validity — independent attestation, and selection that can be reproduced by anyone.
Checkpoint notary
LiveProof that a value existed at a time, recorded by parties who do not work for you.
Committee election
Built — in testingUnbiasable, auditable selection of who sequences — drawn from your own operator set.
Entropy beacon
Live · FreeRandomness whose manipulation costs destroyed value rather than forgone profit.
Checkpoint notary
Live on Arbitrum Sepolia
You file a 32-byte commitment — typically a hash of your own state over some period. It is stamped with the slot open at that moment and frozen when the slot closes. The committee attesting the following slot folds your checkpoint into a value they each derive independently, and two of three must arrive at the same answer before it counts.
What you get is not TRITO's opinion of your data. It is evidence that three parties, selected by a process none of them controlled, recorded the same value at the same time — and that anyone can recompute the selection afterwards to check it was not arranged.
How it is priced: a refundable bond to register, which is collateral rather than a fee and exists so filing costs something. Then a charge per checkpoint, most of which goes to the operators whose machines did the corroborating.
Who does the corroborating
Whichever committee TRITO's own sealing already selected — not a separate pool. Sealing a slot includes folding in any checkpoints filed against it, which are public data already stored on-chain. That is the point rather than a shortcut: the parties corroborating your checkpoint are the ones securing TRITO, drawn by the same randomness, with the same resistance to anyone stacking the committee. A separate opt-in pool would inherit none of that.
What this does not claim
TRITO's committee has not executed your chain and cannot tell whether a checkpoint is true. A fraudulent state root would be corroborated exactly as faithfully as an honest one. What is proven is existence, timing and independent agreement — never correctness. Any service claiming more than that is claiming to have verified something it never saw.
Committee election
Deployed on testnet — first epoch elected and paid; no customer chain has run on it yet
Most rollups today run a single sequencer and intend to decentralise later. The hard part is not running more machines — it is choosing which one produces, in a way that nobody can influence and everybody can check.
TRITO elects a committee from your operator set using Chainlink randomness, weighted by reputation those operators earned over months rather than bought. Each epoch is elected while the previous one is still running, so the schedule is known before it begins. Within an epoch, turns rotate; if the operator on duty goes quiet, the next becomes eligible after a grace period — no accusation, no liveness oracle, nothing for anyone to report or dispute.
Operators are paid from what you pay, never from newly minted TRITO. A fixed sum per epoch served, and a share of a pool drawn by the same randomness that picked the committee.
How it is priced: you prepay into your own escrow and it is drawn down per epoch, scaled by committee size. Stop paying and elections stop — you are buying operator labour, which can be withheld, not access to data, which is public and could not honestly be metered.
What an operator needs: serving a rollup means running your node software alongside their TRITO node — two clients, not one. Rollup clients are mostly not built for phones, so this is desktop or server work today. Operators sealing TRITO from a mobile app earn normally and are not eligible for election work; that is a limit of what rollup software currently runs on, not of the operator. If a customer arrives with a mobile-capable client, those operators become eligible the same day.
What this does not claim
This decides who may produce, not whether what they produced is valid — your own nodes already enforce that. It provides selection provenance rather than on-chain corroboration of your chain's contents; a tenant wanting the second buys the notary alongside. The contracts are live on testnet and one epoch has been elected and paid; the first customer will still be the first production use.
TRITO does not verify that an operator can run your software, because nothing on-chain can observe that. Qualification measures conduct on TRITO — uptime, selections, stake held over time — and none of it proves a machine can host your client. What closes that gap is your own readiness signal: your monitoring confirms your software answered on a given machine, and only operators you have confirmed enter the draw. The check is yours because you are the only one who can make it.
Entropy beacon
Live · Free
Every sealed slot publishes a value binding a proof-of-history hash, an entropy component derived from TRITO actually burned, and the VRF word. Any contract may read it, and a published helper confirms that a seal genuinely binds the components it claims.
What is unusual: most randomness can be nudged by a party willing to give up a reward. Influencing this one requires destroying TRITO. The cost of manipulation is denominated in burned value rather than forgone profit.
Why it is free: it is published on-chain and could not be metered without pretending to withhold something already public. Charging for it would be charging for a restriction we cannot impose.
What this does not claim
TRITO does not generate randomness. It consumes Chainlink VRF and mixes in burn-derived entropy — defence in depth, not a better primitive, and a consumer who already trusts Chainlink completely gains little. It is also slot-paced, so it suits a weekly draw or a jury selection, not randomness on demand.
A note on what is being sold
Published on-chain state cannot be paywalled. Anyone may read the selection, the seals and the weights, and TRITO could not tell who did. What can be charged for is labour — operators running machines, attesting, and being available — because that can be withheld.
This is why the beacon is free and the other two are not, and it is a structural fact rather than a pricing decision. Any part of this that was metered would be metered dishonestly.