On-chain architecture & trust
A prop firm asks a trader to trust a lot: that the fills were fair, that the rules didn't move, that a payout won't get stuck in "under review" forever, that the firm has the money it claims to. DecentralProp's answer to each of those is a specific on-chain mechanism, not a policy promise.
Five programs, one protocol
The protocol runs on five Anchor programs on Solana:
| Program | What it does |
|---|---|
firm | A firm's identity, treasury, fee splits, and token supply |
bonding_curve | The token market each firm's payouts draw from |
challenge | Evaluation and funded-account lifecycle, rules snapshots, settlement |
batch | Hourly commitments of trade history into verifiable roots |
dispute | Fraud challenges and operator-stake slashing |
Every firm on the platform runs on the same five programs. There's no per-firm fork that could quietly behave differently.
The money sits where nobody holds a key
Before any of the mechanisms below matter, one question has to be settled: a firm's treasury holds the funds your payout comes from, so who can move them?
The answer is nobody, in the literal sense. A firm's treasury, its payout staging account, its backstop escrow and its reserve are all program-owned accounts. On Solana these addresses are built so that no private key exists for them at all: an ordinary Solana address is a public key, and these are deliberately constructed to be addresses that no key can produce. There is no seed phrase for a firm's treasury in a safe somewhere, because there is no seed phrase for it anywhere. Funds leave only when the program's own rules say they leave.
That is what separates "the operator won't take the treasury" from "the operator can't." It is the foundation the rest of this page is built on, and every other protection here assumes it.
Trade history can't be quietly edited
Closed trades feed into an hourly settlement process: the trade evidence is assembled into a cryptographic commitment (a two-root transcript covering both the trade steps and the price feed they were priced against), and that commitment is what gets proposed on-chain. If a firm's reported outcome doesn't match what its own trade history actually shows, that mismatch is provable, not a matter of taking the firm's word for it. A watchtower process exists specifically to check every proposed settlement against the real transcript during its challenge window.
Figure — from a closed trade to a finalized (or faulted) settlement
Trade closes
A closed position becomes trade evidence.
Hourly batch commit
Trade history hashes into a verifiable on-chain batch root.
Propose
The settlement, and its payout claim, is proposed on-chain.
Challenge window
A watchtower checks the proposal against the real transcript.
Finalized
Matches the transcript. Payout proceeds normally.
Faulted
Doesn't match. The operator's bond is slashed.
Payouts settle on a timer that scales with trust
A funded withdrawal isn't held for a flat, arbitrary window. A firm paying out of its own solvent treasury settles in well under an hour; the window only extends toward its maximum when a firm is stressed enough to be leaning on the shared backstop layers, exactly the case where extra scrutiny is warranted. Payouts are never cancelled outright, only queued, they release automatically once the check clears.
Cheating costs the operator their own stake
Every firm posts a 50 SOL collateral bond, held in the dispute program. Notably, an operator never has to front this out of pocket, it fills itself automatically from a small slice of the firm's own evaluation revenue over time. If a settlement is ever proven fraudulent, the entire bond is seized: half goes to whoever caught it, half to the protocol's dispute fund. This is what makes "the firm can't fake a result" a real economic claim rather than a policy, cheating costs the operator capital, and getting caught pays the person who caught them.
Going quiet isn't a safe alternative to cheating
An operator can't dodge all of this by simply never settling a passed evaluation, stalling forever instead of risking a provable lie. If a challenge sits unsettled long enough past its deadline, the trader can open a dispute without needing the usual evidence a settlement would normally require, and if the operator still hasn't resolved it after that, the trader is paid directly from the insurance fund. There's no version of this where doing nothing is the winning move. See How settlement actually works for the full fault-proof mechanics, including the specific proof for each way a firm might try to cheat.
The whitepaper sets all of this out as seven numbered trust layers, and is explicit in §6.8 about the assumptions that remain rather than claiming there are none.
Solana is where this runs today, not the only place it can. See Built to run on any chain for why the five programs above are a chain adapter rather than the foundation, for the proof on DecentralChain's own mainnet, and for how the keyless-treasury guarantee above is reproduced on a chain built on completely different foundations.
They can fake or manipulate your fill prices
Pyth-signed prices
Every fill carries an attestation from 90+ independent publishers. Prices can't be faked.
Rules change after you start — moving goalposts
Rules locked on-chain at purchase
Your rule snapshot is immutable for the entire life of your challenge.
Trades deleted or false breaches claimed
Hourly Merkle roots + signed receipts
Your trade history can't be altered or omitted — and you can dispute it on-chain.
Payouts stalled, denied, or 'under review' forever
Guaranteed on-chain settlement
Payouts are never cancelled, only queued. Force-resolution after 7 days; an insurance fund backs disputes.
You never know if the firm is solvent
Transparent treasury + Autonomous Risk Engine
Firm health and reserves are visible on-chain in real time.
They can rug-pull and vanish with your money
Treasury in an on-chain PDA — no backend keys
An operator stake and a fixed-supply token mean no one can drain or mint.
The house wins only when you lose
Positive-sum payouts
Your payout buys the firm's token, so every holder gains. Aligned, not adversarial.
