Am I allowed to move this — and could I actually get out of it?
Compliance tooling reads account flags and never opens the order book. Market terminals read the book and never check the flags. NOSHASHI answers both from the same validated ledger state, in the same second, and returns GO/HOLD/NO-GO with the exact rule that decided it.
Free tier · no account required to start · nothing is ever signed
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 asset is two facts, not one
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
Who can take this away from you?
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
Would anyone buy it if they didn't?
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
Book depth is mostly fiction
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
One ledger read, two questions, one receipt
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
Check it before you pay it
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
Published facts
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
No invented reputation
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
A lookup, not a wallet
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
Built only on what the network has actually activated
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
Credentials, domains, permissioned DEX
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
Freeze and recovery posture
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
Travel Rule and concentration
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
What a payment actually moved
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
How few people can move a treasury
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 · What the ledger will let you believe
Every reading here exists because a field lies
Not edge cases. These are the default readings — what you get if you take the obvious field at face value — and each one is handled explicitly, documented at the top of the module that handles it. The figures are measured on mainnet, not modelled.
delivered_amount ≠ Amount
A tesSUCCESS 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.
250×Over-credit from reading the requested figure. One real transfer asked 999,332.87 and delivered 3,958.64.
Quoted depth ≠ fillable depth
An offer rests in the book whether or not its owner still holds the asset. Summing the advertised figure counts exits that cannot happen.
92.8%Of one book's visible depth could not fill. A single offer advertised 1,400,100 USD against a 22,273 balance.
Sequence is not a count
Since DeletableAccounts a new account's sequence is seeded to its creation ledger index, so it describes when the account began, not what it has done.
92,835,117Read from a live AMM account that has sent zero transactions. The obvious reading is wrong by the whole number.
Quorum is weight, not headcount
XRPL compares a multi-signature quorum against the sum of signing weights. Counting signers describes a committee that may not exist.
5 → 1Five signers where one carries the quorum is a single-key account. We report the minimum who must actually agree.
Absent ≠ false
A trust line carries freeze and authorized only when set. Reading absence as "not frozen" asserts a guarantee the ledger never made.
not statedWhat we report instead. An unset flag is unknown, and unknown is not the same as safe.
Ripple epoch ≠ Unix epoch
Escrow times count from 2000-01-01, not 1970-01-01. The two are both plausible integers and only one is right.
30 yearsHow early every release date reads if the offset is missed. Silent, and wrong in the safe-looking direction.
Thirteen are documented in full, with the module that handles each, in the repository README.
07 · Network reality
A feature exists in three states and only the third one works
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.
ACTIVATED — BUILT ON
NOT ACTIVATED — NOT OFFERED
01
Read
Validated mainnet state, with four public nodes compared and their disagreement reported rather than averaged away.
02
Adjudicate
A deterministic rule set. The same state and the same rules return the same verdict, on your machine and through the API.
03
Digest
SHA-256 over the verdict and every check id and result. Byte-stable, so a receipt can be re-verified years later.
04
Hand over
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 · XRPL Edge Lab
New questions for a public ledger
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.
Download the lab
Get the portable manifest, replay checklist and data dictionary for controlled review. The pack contains no signing keys and cannot submit transactions.
Open download systemLiquidity escape map
Trace 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 · EXPERIMENTALAmendment drift radar
Compare 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 PREDICTIONRecovery graph
Surface 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.
09 · Download
Take the instrument offline
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 · 3.3 MBe144ab0b955494b4c5c9389da3afa4033280cb8432f1e3e1a0829a052bb974ef
On an Intel Mac? Download the Intel build — 3.4 MB, sha256 2ce31ffd…90d8326e. Not notarised, so Gatekeeper will warn: right-click › Open.
Windows
10 and 11 · x64
Built on a Windows runner
Download .exe · 1.7 MB · v0.2.2689089ae5910b0e1e3b32e61f7c1519f4e3d02d4593e8e561b18679ca63f3dc9
Deploying it? The .msi is 2.2 MB, sha256 52baae06…03e496a4. Neither build is code-signed, so SmartScreen will warn.
Linux
x64 · .deb, .rpm, .AppImage
Choose the package for your environment
Download .deb · 2.4 MB · v0.2.231f5ccddced8164bc963b8025cb4be6fbab09ad3aed9a2ae7080ae9a6763f420
Also .rpm (2.4 MB, sha256 b79846f0…f53eafbe) and .AppImage (76.7 MB, sha256 01372e31…33847992). 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_0.2.2_aarch64.dmg # macOS / Linux certutil -hashfile NOSHASHI_0.2.2_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 235 tests. Source on GitHub →
WHAT YOU GET FREE
- Live XRPL mainnet telemetry
- GO / HOLD / NO-GO adjudication, unlimited
- Cryptographic receipts
- Credential registry and domain grid
- 10 address checks per month
- Animated explainers
- On-device compliance agent
- Exit liquidity, freeze-rights and control-surface analysis
- Persistent adjudication ledger
- Travel Rule scoping and signed export
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.
09 · Pricing
Free where a human decides
The console is free forever. Paid tiers begin where a desk needs something an individual does not.
FREE
Free
forever
Individuals and single desks
- Full console and menu bar HUD
- Ledger sync — four public nodes compared
- Inbox — spot impersonated tokens sent to you
- Token rights — check an NFT before you buy it
- Unlimited GO / HOLD / NO-GO gate checks
- 10 address checks per month
- On-device compliance agent
- CSV audit export
- Binary integrity verification
PRO
$749
per seat / month
Trading desks and funds
- Everything in Free
- Unlimited address checks
- Exit liquidity — freeze risk × depth × concentration
- Settlement forensics — delivered vs requested
- Order book integrity — quoted vs fillable depth
- Counterparty provenance — age and funding source
- Treasury control surface — signers and locked value
- AMM pool governance — fee votes and auction slot
- Issuer freeze-rights and XLS-77 deep freeze analysis
- Counterparty concentration (HHI)
- Persistent adjudication ledger — 10,000 verdicts
- Wallet explorer and editable policy rule set
- Issuer drift monitor with native alerts
- 5,000 API verifications · priority support
INSTITUTIONAL
$4,000
per month
Regulated venues and custodians
- Everything in Pro
- Issuance surveillance — holder concentration
- Travel Rule (FATF R.16) scoping
- Signed audit export — SHA-256 chain of custody
- Offline adjudication on segregated networks
- Compliance API and webhooks
- White-labelled wallet · regulator read-only seats
- 100,000 API verifications · published SLA
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.
11 · Security posture
The smallest possible attack surface
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.