NOSHASHI
AI-powered intelligence for decentralized financial ecosystems, with the evidence behind every answer.
The pond answers each validated XRPL ledger as it closes.
From ledger to decision
See the ecosystem.
Every validated ledger as it closes, read from public XRPL servers. Nothing is read from a ledger that can still change.
Watch it liveIdentify meaningful signals.
Concentration, velocity, failures, issuer powers and funded order-book depth, measured from ledger state rather than estimated.
See the signalsTranslate complex network behavior.
Each result names its rule, its threshold and the ledger it was read at. The AI assistant explains results in plain language and cites them.
How it decidesExplore entities and relationships.
Trace source of funds across hops, open a case from any verdict and keep a hash-chained case log your team shares.
Institutional toolsSupport informed professional workflows.
GO, HOLD or NO-GO against your own versioned policy, with a SHA-256 receipt an auditor can recompute. People decide.
The three questionsA decentralized ledger behaves like a garden. Accounts appear, relationships form, clusters bloom around markets and signals travel through them. This is the XRP Ledger now, read in your browser from its latest validated ledgers.
Reading the latest validated ledgers…
The graph appears when this browser has read a validated XRPL ledger. Nothing is drawn from stored or sample data.
Up to 80 of the most active entities are drawn (42 on small screens). Order-book and AMM activity has no counterparty field, so it gathers at a hub. Addresses are shown as the ledger records them; this page does not say who owns them.
An assessment is only as good as the trail behind it. Every line below comes from the same ledgers as the network above, and every one can be checked.
Reading the latest validated ledgers
Asked of a second server after the first read.
The rule is shown here with the first result.
Five kinds of question, one source of truth. Each one is answered from validated ledger state and leaves a record.
Understand the signals that matter to a position: issuer freeze and clawback powers, reserve headroom, concentration and the depth that can actually fill an exit.
In the app: Decision Card, exit simulation against funded depth
Explore who an address deals with, how concentrated that is, and whether its funds touch sanctioned or reported addresses.
In the app: deposit and withdrawal screening, multi-hop source of funds, exchange attribution
See how the ecosystem is structured and what changes from ledger to ledger: clusters, hubs and the amendments actually in force.
In the app: Ledger Watch event feeds, amendment status read from the ledger
Compare observed activity with the conditions you configure, and see exactly which rule held and which blocked.
In the app: versioned policies with four-eyes activation, GO, HOLD or NO-GO with receipts
Turn ledger activity into findings you can cite, with the ledger index and hash attached to each one.
In the app: investigation cases with hash-chained logs and exportable reports. Published findings
NOSHASHI supports compliance work. It does not certify compliance and does not replace your own review.
One ledger. Three questions.
Every answer is read from validated XRPL state, decided by deterministic rules and your versioned policy, and recorded with a receipt anyone can recompute. People make the institutional decision; NOSHASHI shows them what the ledger supports.
COMPLIANCE
GO, HOLD or NO-GO against your policy: freeze rights, credentials, reserve headroom, concentration, Travel Rule thresholds. Every rule's result and reason is shown, and no GO is issued unless independent public servers agree on the ledger.
LIQUIDITY
What the order book quotes is not what can fill. NOSHASHI counts only what each offer's owner can deliver, simulates the exit against that, and joins it to the issuer's right to freeze.
EVIDENCE
Each decision carries a SHA-256 receipt over the rules, the policy version and the
ledger it was read at. Re-verify it, replay it at that ledger, or export a report an
auditor can check with shasum.
LEDGER → VALIDATE → ANALYZE → ADJUDICATE → EXPLAIN → PROVE → APPROVE
One validated state, one deterministic rule set, one receipt an examiner can replay.
Flags, trust lines, credentials, domains, book depth and amendments.
The exact rule, evidence and timestamp behind the answer.
Read-only by design. No custody, key storage or broadcast path.
Price is the midpoint of the XRPL DEX order book for XRP against RLUSD, counting only offers their owners can actually honour. It is the ledger's own price, not a feed from a centralised exchange, and the two can differ.
One validated state becomes a decision an examiner can replay.
Byte-stable output for review and retention.
No keys, custody, transaction builder or broadcast path.
Run the instrument beside the workflow it informs.
01 · The thesis
An issued balance is only an asset if the issuer cannot immobilise it and somebody will buy it. Either alone is a half-answer.
THE COMPLIANCE HALF
XRPL publishes freeze rights in account flags that no wallet reads. We read every issuer behind every position and report which can freeze you out, which have permanently surrendered the right, and — via XLS-77 deep freeze — which can stop you receiving as well as sending.
THE MARKET HALF
We walk the live DEX book and the AMM pool for every position you hold and compute what a full exit would actually achieve — fill rate, price impact, and the depth genuinely reachable rather than the depth nominally resting.
02 · Why the second half is hard
Summing every resting offer counts bids at a thousandth of the market as though they were an exit. Measured on mainnet, GateHub USD/XRP:
| Metric | Naive | NOSHASHI |
|---|---|---|
| Spread | −18,675 bps | 20 bps |
| Full-exit slippage | 9,340 bps | 10 bps |
| Usable bid depth | 12,692,991 USD | 13,894 USD |
The book advertised 12.7 million USD of depth. 13,894 is reachable — a 913× overstatement. The touch was poisoned by a single stale offer priced at 29× the real market; the rest was lowball bids nobody would ever fill against.
03 · How a verdict is made
Both halves are answered from the same validated ledger state, in the same pass. That is the whole architecture — and it is why the two answers can be trusted against each other rather than compared across two tools reading at two moments.
SAME STATEBoth halves read one validated snapshot. A freeze right found at ledger n is weighed against depth measured at ledger n, not against a book that has since moved.
NO SIGNING PATHThe engine is pure. It takes state and returns a verdict — it cannot hold a key, build a transaction or broadcast one, because none of that exists in the build.
BYTE-STABLEThe receipt digest covers the verdict and every check id and result. Restyling cannot change it; a changed digest would invalidate every receipt ever issued, so the engine is fixed by test.
04 · For everyone
Three tools that read mainnet from this page: who you are paying, what a payment actually delivered, and how much of a book could really fill. Free, no account, nothing is signed — and each re-measures a finding we publish, so you never have to take the number on trust.
Can this account freeze what it issues you, and what does it charge?
Try
A payment succeeded. Did it deliver what it said it would?
A tesSUCCESS payment can deliver a fraction of its stated amount. Crediting the stated figure is the mistake that has drained exchanges.
How much of this book could actually fill?
An offer rests in the book whether or not its owner still holds the asset. rippled reports the difference and most interfaces sum the advertised figure anyway. How this was measured.
What are you holding, and who can freeze it?
Every token balance sits on a trust line to whoever issued it, and most issuers keep the right to freeze that line. This reads the lines and then asks each issuer whether it has given the right up.
Who issued this NFT, and what did they keep the right to do?
An XLS-20 token id is not a lookup key — the issuer, the royalty and the rights are encoded in the id itself. This decodes it in your browser, then confirms the issuer exists on mainnet.
Do the public nodes agree on what the ledger says?
You are reading the ledger through somebody's server. If two of them disagree about the validated ledger, that disagreement is the reading — not something to average away.
WHAT IT TELLS YOU
Whether they can freeze your balance, what they charge to transfer, what they have issued and how much is outstanding, what credentials they hold, and who they have actually dealt with.
WHAT IT WILL NOT DO
There is no bad-actor list behind this and no reputation score, because we do not have one. Inventing a number people act on would be the worst thing this product could do.
WHY IT IS SAFE
It moves no money, holds no key and signs nothing. That is deliberate: it keeps a public utility a public utility, with none of the weight of a custodian or a ramp.
05 · For institutions
XRPL ships features as amendments, and a feature exists in three states: specified, implemented in rippled, and activated by validator majority. Only the third one works. NOSHASHI reads the ledger's own amendments object and refuses to offer anything that is not genuinely live.
XLS-70 · XLS-80 · XLS-81
Adjudicate every settlement against a domain's published rule set. Walled gardens where only holders of named credentials may transact, and order books restricted to an approved membership.
XLS-77 · CLAWBACK
Deep freeze blocks a sanctioned holder from sending and receiving — an ordinary freeze leaves them able to keep accepting funds. We read both, plus issuer clawback rights on your positions.
FATF R.16 · HHI
Scope every transfer against your jurisdiction's threshold, and score the settlement book with the Herfindahl-Hirschman Index — the same measure a regulator uses.
DELIVERED ≠ REQUESTED
An XRPL payment carries the amount it asked to deliver and, separately, the amount that arrived. On a partial payment those differ and the transaction still succeeds. A real mainnet transfer we read asked for 999,332.87 and delivered 3,958.64 — 0.4% — returning tesSUCCESS. Credit the requested figure and you are out by two hundred and fifty times. This is the mistake that has drained exchanges, and it is one field.
SIGNER WEIGHT · NOT HEADCOUNT
XRPL compares a multi-signature quorum against the sum of signing weights, not a count of signers. Five signers where one carries the quorum is a single-key account wearing a committee's clothes. We report the minimum number who must actually agree, which is the only number that means anything.
06 · The ledger can be transparent and still be misread
Six real mainnet replies, each read two ways: the obvious field at face value, and what NOSHASHI reports. The right-hand column is not written for this page — it is what NOSHASHI's own code returns for the recorded reply, and a test fails the build if the two ever disagree.
delivered_amount ≠ Amount
A successful (tesSUCCESS) payment carries the amount it asked to deliver and, separately, what arrived. With the partial-payment flag set they differ and the transaction still succeeds. Crediting the Amount field would over-credit by 1,295,975×. This one is an account paying itself through the exchange; the same flag on a payment to an exchange is how deposits get over-credited.
Ledger 107,194,530 · transaction 307662378F…
What a basic interface sees
Amount field
100,000 XRP paid
What NOSHASHI verifies
delivered_amount
0.077162 XRP arrived — 0.00008% of the stated amount
Quoted depth ≠ fillable depth
An offer stays in the book whether or not its owner still holds the asset, and an expired offer is removed only when something touches it. Summing the advertised figures counts exits that cannot happen.
Ledger 107,194,532 · 60 USD.Bitstamp offers for XRP rvYAfWj5gh67oV6fW32ZzP3Aw4Eubs59B
What a basic interface sees
Sum of offers
128,309.49 USD on offer
What NOSHASHI verifies
Funded and unexpired
42,328.77 USD can fill — 33% of what is shown; 31 of 60 offers cannot fill at all
Sequence ≠ transaction count
Since the DeletableAccounts amendment, a new account's Sequence starts at the ledger index it was created in. It says when the account began, not what it has done.
Ledger 107,194,965 · account rHUpaqUPbwzKZdzQ8ZQCme18FrgW9pB4am
What a basic interface sees
Sequence
86,795,414 transactions sent
What NOSHASHI verifies
Sequence − creation ledger
0 transactions sent. The account was created in ledger 86,795,414, and as an AMM pseudo-account it cannot send any.
Quorum ≠ signer headcount
XRPL compares a multi-signature quorum against the sum of signing weights, not the number of signers. The number that describes control is the fewest signers who can reach the quorum.
Ledger 107,194,532 · signer list owner rLdhU5DKrpztzVPnQmtzAbHfiHR8LPYtaX
What a basic interface sees
Signers listed
3 signers control the account
What NOSHASHI verifies
Minimum who must agree
2 of 3 can sign: quorum 2 against weights 1, 1, 1
Absent ≠ false
The ledger answered actNotFound: there is no account at this address. That is not an account holding nothing; a payment to it would have to create it, and nothing about it can be verified.
Ledger 107,194,798 · address rPGJe8F1FzegsNDJEMkq3hUrupovdBjQMi
What a basic interface sees
Balance
0 XRP
What NOSHASHI verifies
Account activated on mainnet
FAIL — Address is well-formed but has never been funded on mainnet.
Ripple epoch ≠ Unix epoch
Ledger times count seconds from 2000-01-01, not 1970-01-01. Both readings are plausible dates and only one is right; the wrong one is 30 years early.
Ledger 107,194,530 · transaction 307662378F…
What a basic interface sees
date as Unix time
1996-09-24 02:42:12 UTC
What NOSHASHI verifies
date from 2000-01-01
2026-09-24 02:42:12 UTC
Each case links the object it was read from on the public explorer. The same cases, with the module behind each reading; thirteen traps in all are documented in the repository README.
07 · Network reality
XRPL ships features as amendments: specified, implemented in rippled, and activated by validator majority. Until the third, every transaction of that type is rejected by every validator on the network. NOSHASHI reads the ledger's own amendments object and computes each id locally — SHA-512Half of the feature name — so no lookup table can drift, and anything not activated is never offered.
01
Validated mainnet state, with four public nodes compared and their disagreement reported rather than averaged away.
02
A deterministic rule set. The same state and the same rules return the same verdict, on your machine and through the API.
03
SHA-256 over the verdict and every check id and result. Byte-stable, so a receipt can be re-verified years later.
04
Signed export with a chain of custody an examiner can follow, plus the request id that names the same adjudication in your logs and ours.
Amendment state verified against mainnet on 2026-08-24 — rippled 3.3.0, 93 amendments active. The app re-reads it live; this list is the state at the last audit of this page.
08 · XRP, live
Price and seven days of history, the validated ledger the network is on right now, and headlines from three public feeds — all read on the server before this page was sent. Context, not signal: nothing in this section is a NOSHASHI reading, and no verdict follows from any of it.
Where has XRP traded this week?
How much actually changed hands?
Merged from Google News, Cointelegraph, CoinDesk. Headlines link to the publisher and are reproduced as titles only. They are news about the XRP Ledger, not a NOSHASHI reading, and no verdict is implied by any of them.
09 · XRPL Edge Lab
Three experimental analyses built from validated XRPL state. They are designed to expose operational blind spots, not to predict markets or replace an accountable review.
Get the portable manifest, replay checklist and data dictionary for controlled review. The pack contains no signing keys and cannot submit transactions.
Open download systemTrace the shortest credible path from a held issued asset to an XRP exit, including issuer controls, trust-line constraints and observed book depth.
READ-ONLY · EXPERIMENTALCompare the active amendment set with the rule assumptions used in an evidence receipt, so a previously valid control cannot quietly age into a false comfort.
REPLAYABLE · NO PREDICTIONSurface issuer, signer-list and credential dependencies that can change who is able to remediate a blocked position. Unknown data stays unknown.
PUBLIC DATA · REVIEW REQUIREDExperimental means the rule definitions and coverage may change. Do not use an Edge Lab output as the sole basis for a trade, filing, customer decision or legal conclusion.
10 · Download
The full build, free to download. Bring the same adjudication engine to a controlled workstation, then verify the artifact before it ever touches a desk.
macOS
11 Big Sur or later
Apple silicon and Intel built separately
Download .dmg · Apple silicon · 6.2 MB · v1.0.20bbefb15da299c25ee040f045c80d8e19cd026413d4b4aece71917dda85c70a84
Intel build (.dmg) — 6.4 MB, sha256 9b8d1362…aa47d2f9. Not notarised, so Gatekeeper will warn: right-click › Open.
Windows
10 and 11 · x64
Built on a Windows runner
Download .exe · 3.2 MB · v1.0.20a4a8ea5b588d887420b629f7302100a7188056a6a6cb41b4b16fe15f79eceb49
Deploying it? The .msi — 4.0 MB, sha256 f7da8d8a…95e4fe07. Neither build is code-signed, so SmartScreen will warn.
Linux
x64 · .deb, .rpm, .AppImage
Choose the package for your environment
Download .deb · 4.4 MB · v1.0.2066ce5a67e60857226884185912bb8c889a87096468e208c1bf835e81511ce4d5
.rpm — 4.4 MB, sha256 28a3b46f…aac4a9e0 and .AppImage — 83 MB, sha256 29fc0280…badfa9a6. The AppImage is large because it carries its own WebKitGTK; the .deb uses yours.
VERIFY BEFORE YOU RUN IT
Check the download matches the hash above:
shasum -a 256 NOSHASHI_1.0.20_aarch64.dmg # macOS / Linux certutil -hashfile NOSHASHI_1.0.20_x64-setup.exe SHA256 # Windows
The app also hashes its own running binary — Settings › Binary integrity — so you can confirm nothing changed between this page and execution. That check is free on every tier; charging for the ability to verify we are not malicious would be perverse.
You can go further and read the code. Every build on this page was produced by CI from a public commit, and the logic that makes claims carries 272 tests. Source on GitHub →
WHAT YOU GET FREE
Not code-signed or notarised yet. macOS will warn about an unidentified developer — right-click the app and choose Open the first time. Better said here than as a surprise.
11 · Mission log
Releases reach this log from the GitHub Releases API, which means an entry cannot appear without CI having produced an artifact first. Maintenance and incidents are written in the repository and reviewed like code. Neither is typed into a box on a marketing page.
FLIGHT READINESS
RECENT ENTRIES
Built from c310861dcc77d9c81b7d31482b9e52410df3b7a0 on macos-latest. Verify your download before running it: shasum -a 256 # macOS / Linux certutil -hashfile SHA256 # Windows The app also hashes its own running binary — Settings › Binary integrity — so you can confirm nothing changed between this…
RELEASE NOTES →Built from 10c43706994495324049249727adb1c0c48872cc on macos-latest. Verify your download before running it: shasum -a 256 # macOS / Linux certutil -hashfile SHA256 # Windows The app also hashes its own running binary — Settings › Binary integrity — so you can confirm nothing changed between this…
RELEASE NOTES →Built from 0e8ff5050fc19a40ceff02c34c68446c1416990d on macos-latest. Verify your download before running it: shasum -a 256 # macOS / Linux certutil -hashfile SHA256 # Windows The app also hashes its own running binary — Settings › Binary integrity — so you can confirm nothing changed between this…
RELEASE NOTES →Built from 2fad8055b3a5fecc0b741caec8bfe8223e8db20b on macos-latest. Verify your download before running it: shasum -a 256 # macOS / Linux certutil -hashfile SHA256 # Windows The app also hashes its own running binary — Settings › Binary integrity — so you can confirm nothing changed between this…
RELEASE NOTES →Built from b8c9b2c80cb5bad57249c22633f00d858c01cb44 on macos-latest. Verify your download before running it: shasum -a 256 # macOS / Linux certutil -hashfile SHA256 # Windows The app also hashes its own running binary — Settings › Binary integrity — so you can confirm nothing changed between this…
RELEASE NOTES →PRODUCT UPDATES
An email when a build ships or a capability lands — drawn from the same mission log this page publishes, so an email cannot claim something the site does not. No schedule, no digest, one click to leave.
12 · Pricing
The console is free forever. Paid tiers begin where a desk needs something an individual does not.
FREE
Free
forever
Individuals and single desks
PRO
$749
per seat / month
Trading desks and funds
INSTITUTIONAL
$4,000
per month
Regulated venues and custodians
ENTERPRISE
$10,000
per month
Institutional teams and regulated operations
STRATEGIC INFRASTRUCTURE
$20,833
per month
Teams building their own XRPL intelligence layer
Subscriptions renew automatically until cancelled. Cancel any time in two clicks. Full refund within 14 days of a first charge if no paid capability was used. NOSHASHI sells software. It does not transmit money, hold customer funds, or take custody of any asset.
13 · Common questions
The same answers the support console gives, from the same file. If the two ever disagree, this page is right and the console is a bug.
Something not here? The support console, bottom right, answers from this same set — and hands anything it cannot answer to a person.
14 · Security posture
The safest data is data nobody holds. Without an account there is no server-side state at all — no analytics, no telemetry, no crash reporting.
15 · Contact
There is no sales qualification funnel and no account to create first. Say what you are trying to work out and you will get a straight answer.
01 · PRODUCT
How something works, a build that will not open, a reading you do not trust.
02 · DESKS
Deployment on a controlled network, seats, the compliance API, procurement questions.
03 · DISCLOSURE
Vulnerability reports. Send them here before anywhere else and they get acted on.
04 · DATA
What is held, what is not, and what happens to a message sent through the form.
