The Nine XRPL Addresses Behind FXRP
FXRP is XRP that has been locked on the XRP Ledger so a copy of it can be used on Flare, a separate blockchain. The locked XRP is real, so it has to sit in real accounts — and it sits in exactly nine of them. Read at a single ledger, those nine hold slightly more XRP than there is FXRP in existence.
The nine addresses
Seven of them hold the reserve. Six are run by independent operators called agents, who each take XRP from users and hand out FXRP in return. The seventh, the Core Vault, is a shared pot they can all draw on. The remaining two handle fees and escrow payouts and hold no reserve.
| Address | Role | Spendable XRP | Escrowed XRP |
|---|---|---|---|
| Core VaultrfkXSaCZKTg1EZzec2rLDyrWHxRVJdtVXj | A shared pot of XRP that any agent can draw on, so reserves are not stranded in one place.CoreVaultManager.coreVaultAddress() | 6,916,363 | 140,000,000 |
| Bifrost WalletrEUvL6uJ1NYqa81tGxjH6BnLCQRzqT1aR7 | AgentAssetManager.getAllAgents() | 323,968 | — |
| NORTSOrGK7w4oK5x4ncCZyjYEH4QpCZSkKSMrUV2 | AgentAssetManager.getAllAgents() | 257,540 | — |
| Atlas TSOraDZBXEKsBtgcRsWjq6bg2DVRkCXknJ4Kq | AgentAssetManager.getAllAgents() | 221,514 | — |
| AUrU3KcE1fqgn7Qm8dA3DSnefNoUnsYF31Wr | AgentAssetManager.getAllAgents() | 214,858 | — |
| White KnightrGWGTbxqmpLJ3TYGjoCFDqY9mPQ4T1GTGb | AgentAssetManager.getAllAgents() | 192,360 | — |
| Oracle-daemonrs6q5K4RrVbrymieYjx9TFzTW24Rv9BekL | AgentAssetManager.getAllAgents() | 132,796 | — |
| XRPL Provider WalletrM2LEysS4isvAJkZFfxKDL5z4aTfWcBTXV | Collects the fees on instruction payments. Not backing.MasterAccountController.getXrplProviderWallets() | 36,203 | — |
| Core Vault CustodianrMLNvZR9dascY5jtCfCv3whAp8HdUSZAQ | The one account every escrow is set to pay out to. Holds almost nothing itself.CoreVaultManager.custodianAddress() | 5 | — |
| Reserve total (7 addresses) | 8,259,398 | 140,000,000 | |
Nothing in that list is our own guesswork. Each address is returned by a Flare contract when you ask it, and the call is printed beside the address. The set is closed: the contract that registers agents reports exactly 6, so there is no seventh agent to miss.
Most of the money is invisible to a balance check
Look up the Core Vault on any block explorer and it shows about 6,916,363 XRP. Add up the seven reserve balances that way and you get 8,259,398 XRP against 148,161,053 FXRP — which reads as 5.6% backed, and would be alarming if it were true.
It is not, because of how the XRP Ledger handles escrow. An escrow is a payment locked in advance: the XRP is set aside now and can only be collected later, by a named recipient, after a date. The moment an escrow is created the XRP leaves the sender's balance and moves into a separate ledger entry. It is still there and still theirs, but a balance lookup no longer counts it.
The Core Vault has 14 of them, each for exactly 10,000,000 XRP — 140,000,000 XRP in total, or 94.4% of the whole reserve. They all pay to the same place, the custodian account rMLNvZR9dascY5jtCfCv3whAp8HdUSZAQ, which is why that account holds almost nothing itself: it is a destination, not a vault.
A ladder, one rung a day
The expiry dates are the interesting part. They are consecutive — one escrow comes due every single day from 2026-08-30 to 2026-09-12:
That is a rolling ladder rather than one big lock-up. At any moment thirteen rungs are still sealed, so the reserve cannot be moved in a single action, while one rung frees up each day to keep redemptions flowing.
What actually unlocks a rung is worth being precise about, because "no finish-after date" sounds like the custodian can simply take it. It cannot. Every one of the 14 escrows carries a crypto-condition — a lock that opens only for whoever holds the matching secret. The XRP Ledger requires this: an escrow with no finish-after time must carry a condition instead, or the network rejects it outright (rippled EscrowCreate.cpp, lines 158-163, which returns temMALFORMED). A rung can therefore be released at any moment, but only by presenting that secret, and the XRP goes to the custodian either way.
The deadline is a hard stop in the other direction too. Once the cancel time passes the escrow can no longer be finished at all — the attempt fails with tecNO_PERMISSION (rippled EscrowFinish.cpp, lines 245-251) — and the only remaining move is to cancel it, which returns that 10,000,000 XRP to the Core Vault. Each rung is a one-day window: claimed with the secret, or sent back to be re-laddered.
Does it add up?
Counting the escrowed XRP where it actually is, the seven reserve addresses hold 148,259,398.30 XRP against 148,161,053.28 FXRP in circulation, both read at the same moment.
It covers, with 98,345 XRP to spare. That surplus is not a mystery: agents keep the minting fee users pay them, in XRP, in the same account, so a small excess is what a working system should look like. A shortfall would be the finding; there is not one.
Two caveats on what this does and does not prove. It shows the XRP exists and is held by the accounts Flare's own contracts point to at one moment in time — a single ledger, not a guarantee about the next one. The other half of the question is whether an agent could fail anyway: each must also lock up money of its own on Flare — collateral, which it forfeits if it does not pay out. Asking the contract, all 6 are currently marked normal, none in liquidation, each holding between 211% and 598% of the value it has issued against a requirement of full cover. All six charge the same 0.1% to mint. That is the cushion as it stands today — a live number, not a permanent state.
Who actually uses it
Flare's public explorer logs each mint and each redemption as it happens. Since 2025-12-11 it has recorded 54,100 mints and 19,102 redemptions, between them involving 15,801 different XRP Ledger addresses. Those counts are reliable; the XRP totals attached to them are not the same thing as supply, and are not used that way here — see the note at the end of the next section.
Minting is concentrated. One agent has taken more than half of it:
Half of them hold nothing else on the ledger
Every address that has minted, redeemed or sent an instruction can be looked up against the rest of the XRP Ledger, which says something about who showed up. After setting aside the exchange addresses that appear because a redeemer can send the payout straight to one, 15,777 wallets are left, and they are small: the median holds 10 XRP — barely above the reserve every account must keep — with 497 at the 75th percentile and 6,438 at the 90th. This is retail, not institutions.
What they hold beyond XRP is thinner than you might expect. 47% hold at least one token trustline (7,468 wallets, 144,471 lines between them) and 13% hold at least one NFT (2,104 wallets, 143,765 NFTs). The median wallet holds neither: 51% — 8,103 of them — have no token and no NFT at all.
The reason is that FXRP drew on two different crowds. 10,206 of these wallets already existed before FXRP did, and 60.7% of those hold a token — roughly the profile of an established XRP Ledger user. The other 5,246 were opened after launch, and only 20.9% of them do. Nearly all of the hold-nothing half sits in that second group: wallets that came for this and have not touched anything else on the ledger.
The tails are long in both directions — one wallet carries 2,500 trustlines and another 10,228 NFTs — so a handful of heavy users account for a large share of the 144,471 lines and 143,765 NFTs above. That is why the figures here are medians and percentiles rather than averages.
Since 2026-02-17 there is also a second way in. Rather than using Flare directly, a user can send an ordinary XRP payment to the provider wallet with a short instruction written into the memo field, and Flare carries the action out on their behalf. 45,550 of those instructions have been sent. It means someone can mint FXRP, or lend it out on Flare to earn interest, without ever holding a Flare account or any FLR — the coin Flare charges its own transaction fees in.
What a depositor actually gets
Minting FXRP does not pay anything by itself. You hand over XRP and get the same number of FXRP back, minus the agent's fee of 0.1% — no interest, no reward. Whatever a depositor earns comes from what they do with the FXRP afterwards, and the instruction memo lets them choose that in the same payment. There are three places to send it, and Flare's contract lists them by number.
This is where most FXRP has gone: 94,210,885 of the 148,161,053 in existence, 64%, is sitting in these three vaults rather than being held as a coin.
| Vault | FXRP in it | Room left | Earned, annualised |
|---|---|---|---|
| ClearstarUpshift · vault 2 · 35,008 deposits from the XRPL | 33,810,626 | 11,189,374 | 1.29% over 30d · 1.41% over 90d |
| MonarqUpshift · vault 3 · 3,291 deposits from the XRPL | 5,890,168 | 4,109,832 | 7.01% over 30d · 2.73% over 90d |
| Firelight stXRPFirelight · vault 1 · 153 deposits from the XRPL | 54,510,091 | 5,657,909 | none in the token — by design |
Those rates are not advertised numbers. Each one is the vault's own share price today against its share price at a block that many days ago, annualised — what a depositor who put money in then and left it there has actually earned.
The bottom row needs care, because the obvious reading of it is wrong. Firelight stXRP holds more FXRP than the other two together — 54,510,091 — and its value has not moved: the contract's own history goes 1.000042 in December to 1.000072 in May and has been frozen there for four months. That is not a broken vault. Firelight pins stXRP to 1:1 on purpose, so that the token stays interchangeable and usable elsewhere, and it says rewards are "currently paid in XRP" once its insurance model starts charging fees — a second phase that has not begun.
So a Firelight depositor today is earning two things, and neither is interest. One is Firelight Points, a programme that pays out only if it is later converted into something. The other is whatever they make by taking the stXRP and using it somewhere else — and the chain shows plenty of people doing exactly that. Of the 32,927,434 stXRP held by the twenty largest holders, 12,630,255, or 38%, sits inside another protocol rather than a wallet: lending on Morpho, liquidity pools, and Spectra "principal tokens" that let a holder sell the future yield separately from the deposit — one dated 27 August 2026, one dated 26 November 2026.
The catch is who is doing it. The single largest holder is an ordinary wallet, not a protocol, sitting on 8,988,354 stXRP — worth the same again in XRP — that has not moved in 90 days. Its FXRP balance has not moved either. Whoever that is has taken on the wrapped-asset risk and, unless they are counting on points, has earned nothing measurable for three months.
Clearstar, the busiest destination by far with 35,008 deposits sent straight from the XRP Ledger, pays 1.29% — and that rate is falling: it was 1.6% measured over 180 days, 1.41% over 90 and 1.15% over the last week. Monarq is the only one paying meaningfully more, at 7.01% over 30 days, and it is the smallest, holding 5,890,168 FXRP against a 10,000,000 cap.
For context on whether any of it is worth it: 1.29% on Clearstar is the paid return for giving up direct control of your XRP and trusting the agent, the Core Vault and the vault contract. Firelight's answer is instead a points balance and a promise of fees from an insurance business that has not started. Everything earlier in this article describes the machinery a depositor is relying on in both cases.
It already peaked
A wrapped asset only exists while people want it wrapped, so the interesting question is which way the supply is going. Reading the FXRP contract directly every ten days since launch gives the whole history, and it is not a straight line.
FXRP went from nothing to 157,425,535 in about six months, turned on 2026-06-01, and has shrunk every month since. Minting did not merely slow, it nearly stopped: Flare's mint log records 37,188,200 XRP of new mints in March against 2,066,170 in August, a drop of 94%. The falling curve above is the independent confirmation — that one is read from the contract.
So how much FXRP exists in 2050?
There are 8,525 days between this measurement and the first of January 2050. Take the three rates the data actually supports and run each one out that far:
| If you extrapolate… | FXRP / day | FXRP on 1 Jan 2050 | Compared with today |
|---|---|---|---|
| the whole run since launch | +384,472 | 3,425,768,225 | 23× today's supply |
| the last 90 days | -103,123 | nothing left | empty by August 2030 |
| the last 30 days | -67,388 | nothing left | empty by September 2032 |
One reading says 3,425,768,225 FXRP — 23 times everything that exists today. The other two say the supply runs out before 2033. When three defensible readings of the same nine months disagree by everything, the honest conclusion is that this data cannot forecast 2050, and any single number that claims to is picking a rate to suit the answer. What can be said is narrower and firmer: one growth phase ran from December to June, it has ended, and supply has fallen for three months straight.
Two hard limits sit under any figure anyway. FXRP can never exceed the 100 billion XRP that exist. And far tighter than that: before an agent may issue FXRP it has to lock up money of its own on Flare as cover, and it can issue only as much as that cover allows. Across all 6 agents there is room for exactly 846,120 XRP more today. Counting what they already back, the whole agent side of the system can carry 2,074,606 XRP in total — so the top row would need about 1,651 times the cover that exists, on top of a Core Vault the same multiple larger.
Why the reserve is really one address
The agents look like the backbone, and for minting they are. For backing they are not. Asking each agent how much FXRP it currently backs gives 1,228,486 XRP between all 6 of them — against 146,916,363 XRP in the Core Vault. The Core Vault stands behind 99.17% of every FXRP in existence.
That is worth knowing before reading the mint-share chart above as a measure of importance. It also means supply is not really an organic market rate that can be projected — it moves when the Core Vault is funded or drained, which is a decision, not a trend. The contract running that vault names a single account as the only one allowed to trigger its transfers, under Flare's governance contract, so the backing behind 99% of all FXRP is concentrated operationally as well as financially.
A note on method: every supply figure here is read from the FXRP contract itself, never reconstructed by adding up the explorer's mint and redemption feeds. Those feeds log requests and events, which is not the same thing as coins appearing or disappearing, and they do not reconcile with the contract in either direction — subtracting one from the other implies less burning across the whole period than demonstrably happened after June alone. They are cited above only for counting mints, which is what they do record.
Check it yourself
Every address above is returned by a contract call, not compiled by us. Start at Flare's contract registry, which is at the same address on every Flare network, and ask it for AssetManagerFXRP and MasterAccountController. From the asset manager, getAllAgents gives the 6 agents and getAgentInfo gives each one's XRP Ledger address; getCoreVaultManager leads to coreVaultAddress and custodianAddress. The provider wallet comes from getXrplProviderWallets.
Then read the XRP Ledger side at a fixed ledger index — the ledger's sequence number, so every value comes from the same instant and the totals agree: account_info for balances and account_objects with type: "escrow" for the locked rungs. Skipping the second call is exactly the mistake that makes the reserve look like 5.6% of what it is. Compare the total to totalSupply() on the FXRP token contract.
All nine addresses are labelled on xrpl.to, so they are named wherever they appear — the Core Vault included. More in our other insights.