npm package discovery and stats viewer.

Discover Tips

  • General search

    [free text search, go nuts!]

  • Package details

    pkg:[package-name]

  • User packages

    @[username]

Sponsor

Optimize Toolset

I’ve always been into building performant and accessible sites, but lately I’ve been taking it extremely seriously. So much so that I’ve been building a tool to help me optimize and monitor the sites that I build to make sure that I’m making an attempt to offer the best experience to those who visit them. If you’re into performant, accessible and SEO friendly sites, you might like it too! You can check it out at Optimize Toolset.

About

Hi, 👋, I’m Ryan Hefner  and I built this site for me, and you! The goal of this site was to provide an easy way for me to check the stats on my npm packages, both for prioritizing issues and updates, and to give me a little kick in the pants to keep up on stuff.

As I was building it, I realized that I was actually using the tool to build the tool, and figured I might as well put this out there and hopefully others will find it to be a fast and useful way to search and browse npm packages as I have.

If you’re interested in other things I’m working on, follow me on Twitter or check out the open source projects I’ve been publishing on GitHub.

I am also working on a Twitter bot for this site to tweet the most popular, newest, random packages from npm. Please follow that account now and it will start sending out packages soon–ish.

Open Software & Tools

This site wouldn’t be possible without the immense generosity and tireless efforts from the people who make contributions to the world and share their work via open source initiatives. Thank you 🙏

© 2026 – Pkg Stats / Ryan Hefner

@forgesworn/moneyer

v0.5.0

Published

An LNURLcash (LUD-25) mint - strikes Lightning bearer notes. Independent implementation, cln/lnd funding sources, SQLite, zero HTTP framework.

Readme

moneyer

An LNURLcash (LUD-25) mint. A moneyer was the mediaeval craftsman licensed to strike coins; this one strikes Lightning bearer notes.

moneyer is an independent implementation of the LUD-25 draft: paying an invoice it issues mints a bearer note, and a note's holder can rotate, split, merge and melt it against the withdraw callback. The note's spend secret is the invoice's payment preimage, or, better, a secret the buyer chose and named on the way in. It passes the full lnurlcash-conformance grader, including the spending checks, and the grader runs in this repo's own test suite.

Reference stack, by dni (MIT): the LUD-25 draft, lnurl-mint, lnurl-wallet. moneyer shares no code with them; where their behaviour encodes a safety lesson, that behaviour is kept deliberately and tested.

What it is

  • TypeScript, ESM, Node 24+. node:http and node:sqlite; no web framework, no ORM.
  • Funding sources: cln (clnrest) and lnd (REST), both of which accept a caller-supplied invoice preimage - the capability a LUD-25 mint cannot exist without. phoenixd and NIP-47 make_invoice do not offer it, which is why neither can back a mint. A fake backend exists for development and tests and refuses to run outside --dev.
  • Notes are stored by id, sha256(k1) - the store never holds a spend secret. A buyer may name the note they are buying, in which case the secret is theirs alone from the start.
  • Signs every note it mints with its own mint key (secp256k1, the standard Lightning Signed Message construction) for LUD-25 offline verification.
  • The melt discipline: reply OK when the note is reserved, pay in the background, burn only on confirmed payment, restore only on confirmed non-payment, and park everything else as pending for reconciliation - which also runs at startup, so a crash mid-melt never guesses.

Run

Not on npm yet - build from source, with the sibling repos it links against until those publish:

git clone https://github.com/TheCryptoDonkey/lnurlcash-kit
git clone https://github.com/TheCryptoDonkey/lnurlcash-conformance
git clone https://github.com/forgesworn/moneyer
(cd lnurlcash-kit && npm install && npm run build)
cd moneyer && npm install && npm run build && npm run web:build
# development: fake funding source, in-memory store, a funded note printed
node dist/cli.js --dev

# production shape
MONEYER_BACKEND=cln \
MONEYER_BACKEND_URL=https://127.0.0.1:3010 \
MONEYER_BACKEND_RUNE=... \
MONEYER_SIGNING_KEY=<32 bytes hex> \
MONEYER_PUBLIC_ORIGIN=https://mint.example \
MONEYER_DB=/var/lib/moneyer/mint.sqlite \
node dist/cli.js

The mint is then payable at [email protected] (and the bare-domain _ alias).

Configuration

Environment only. Every variable is MONEYER_*; anything unset takes the default, and a variable set to an empty string counts as unset.

| variable | default | what it does | | --- | --- | --- | | MONEYER_HOST | 127.0.0.1 | listen address | | MONEYER_PORT | 3737 | listen port | | MONEYER_PUBLIC_ORIGIN | derived from Host | the origin wallets are told to call back on. Required behind a reverse proxy, and required for zap-to-note | | MONEYER_ONION_URL | | this mint as a Tor hidden service, e.g. http://<v3>.onion. A request arriving on that host gets its URLs built from it (see below) | | MONEYER_USERNAME | mint | the local part of the mint's own lightning address | | MONEYER_DESCRIPTION | an LNURLcash note | what a note is called on the wire (defaultDescription, and description on discovery) | | MONEYER_DB | moneyer.sqlite | SQLite path; :memory: allowed | | MONEYER_BACKEND | fake | cln, lnd, or fake (refused outside --dev) | | MONEYER_BACKEND_URL | | the funding source's REST endpoint | | MONEYER_BACKEND_RUNE | | cln authentication | | MONEYER_BACKEND_MACAROON | | lnd authentication, hex | | MONEYER_SIGNING_KEY | | 32 bytes of hex. Unset means notes go out unsigned, which holders will notice | | MONEYER_PREVIOUS_SIGNING_PUBKEYS | | compressed pubkeys this mint signed under before, comma separated (see below) | | MONEYER_BASE_FEE_MSAT | 0 | flat mint fee | | MONEYER_FEE_PPM | 0 | proportional mint fee, parts per million | | MONEYER_ROUND_FEE_TO_SAT | true | ceiling the mint fee to a whole sat (see below) | | MONEYER_MIN_SENDABLE_MSAT | 1000 | smallest payment the mint advertises | | MONEYER_MAX_SENDABLE_MSAT | 100000000 | largest payment the mint advertises | | MONEYER_MIN_MINT_MSAT | 1000 | dust floor: the smallest note the mint will strike | | MONEYER_MAX_K1S | 21 | most notes one callback may name | | MONEYER_VERIFY | true | the LUD-21 verify endpoint. Off means 404 | | MONEYER_WALLET_URL | | a companion web wallet the mint's site links notes into | | MONEYER_SUNSET | false | wind down: refuse anything that grows liabilities, keep every way out open | | MONEYER_STATS | true | the /stats endpoint. Off means 404 | | MONEYER_STATS_RATIO_ONLY | false | publish the coverage ratio alone, without the size of the book | | MONEYER_STATS_PUBLISH | false | publish a signed hourly snapshot of /stats to Nostr | | MONEYER_ANNOUNCE | false | announce this mint on Nostr, hourly, so a wallet can find it (see below) | | MONEYER_METRICS | false | the /metrics endpoint, in the OpenMetrics text format | | MONEYER_NAME_PRICE_MSAT | | what a self-service lightning address costs. Unset means registration is closed; 0 means free |

Who runs this mint

Optional, all of it, and unset means the field is simply absent from the discovery endpoint rather than present and empty. This is the human layer a holder wants before trusting a mint with sats.

| variable | what it does | | --- | --- | | MONEYER_NAME | the mint's name, shown on its own site and published as name | | MONEYER_CONTACT_NOSTR | an npub or hex pubkey to reach the operator on; published as an npub | | MONEYER_CONTACT_EMAIL | an email address to reach the operator on | | MONEYER_CONTACT_URL | a contact page | | MONEYER_TOS_URL | the terms a holder is agreeing to | | MONEYER_MOTD | a message of the day, at most 280 characters: maintenance, a sunset, a fee change. It is how an operator talks to holders without a mailing list they never signed up to |

The discovery endpoint also publishes fees (the structured twin of the payRequest metadata's fee prose), version, and previousPubkeys.

The website

GET / serves the mint's own site (web/, vite + anime.js, built by npm run web:build into memory-served static files). It is a wallet-grade LNURLcash client in its own right, driven entirely by lnurlcash-kit against the same endpoints every wallet uses:

  • Mint a note in the browser: amount in, fee grossed up and shown before the invoice exists, invoice QR (tap opens a wallet), LUD-21 polling with a countdown, and on settlement the note is claimed and immediately rotated - the preimage any invoice-observer could poll out of verify is dead before the note is shown. The rotated note's signature is verified against the mint's advertised key in front of the user, and the QR arrives under scratch-off silver foil, rubbed away like a scratch card.
  • Check a note: live value, spent/unknown/pending classified in plain words, offline signature verification.
  • MONEYER_WALLET_URL (optional) links minted notes straight into a companion web wallet's #/claim route.

Without a web build the server falls back to a self-contained, zero-script landing page, so an npm-installed mint still has a face.

As a library:

import {createMoneyer, configFromEnv} from '@forgesworn/moneyer'
const mint = await createMoneyer(configFromEnv())

Rotating the signing key

moneyer signs notes with its own key rather than the funding node's, so the node can be swapped without touching a single outstanding note. The signing key itself is the harder one: rotate it naively and every note already out there stops verifying at once, and every wallet that pinned the old key refuses the mint on its next contact.

So the old pubkeys stay published. previousPubkeys on the discovery endpoint lists the keys this mint has signed under before, and a wallet that finds its pin in that list treats the change as a rotation rather than an impostor. The trust argument is TOFU's own: whoever controls the host controls the pin either way, and the history only stops a legitimate rotation from looking like an attack.

To rotate:

  1. Generate a new key: openssl rand -hex 32.
  2. Set MONEYER_SIGNING_KEY to the new key, and add the old public key to MONEYER_PREVIOUS_SIGNING_PUBKEYS (compressed, 33 bytes of hex, comma separated, oldest last). The old private key is not wanted here and should not stay on the server: verifying an old note needs only the pubkey.
  3. Restart. Notes signed under the old key keep verifying against it; every note minted or mutated from now on is signed under the new one, so a holder who rotates their note re-signs it under the new key for free.
  4. Tell holders through MONEYER_MOTD.

Both variables are validated at startup: an entry that is not a compressed point on the curve, or that repeats the current key, stops the mint rather than quietly publishing a key that verifies nothing.

Transparency: what the mint owes

LNURLcash notes are not blinded, so a mint can state its liabilities exactly. No epochs, no blinded sums, no ceremony: the mint knows every note it has issued and what each is worth, and it can add them up.

curl -s https://mint.example/stats | jq
{
  "at": 1755800000000,
  "outstandingMsat": 48120000,
  "outstandingNotes": 12,
  "pendingMsat": 0,
  "pendingMelts": 0,
  "oldestPendingMeltAgeSecs": 0,
  "localBalanceMsat": 92400000,
  "coverage": 1.9202,
  "reconciledAt": 1755799800000
}

coverage is the funding node's outbound balance over the mint's outstanding liabilities, to four decimal places. Under 1 means the node could not pay every note out today; that is the operator's business to show, not the endpoint's to hide. A mint with nothing outstanding has no coverage figure at all, because "infinitely covered" is not a claim worth making, and a funding source that will not report its balance leaves both localBalanceMsat and coverage out rather than have them guessed.

The figures are cached for 30 seconds, are never per-note, and appear as one row on the mint's own site and its fallback landing page. The endpoint is public by design. MONEYER_STATS=false switches it off; MONEYER_STATS_RATIO_ONLY=true publishes the ratio without the size of the book.

Signed snapshots

With MONEYER_STATS_PUBLISH=true, and the Nostr identity that zap-to-note uses, the mint publishes an hourly snapshot as a parameterised replaceable event (kind 30078, d tag lnurlcash-liabilities). Its content is the /stats JSON plus a sig over the canonicalised (RFC 8785) body, made with the note signing key - the key a holder already checks their own notes against, so the history adds nothing new to trust.

node scripts/verify-stats.mjs <mintPubkey> snapshot.json

Announcing the mint

A wallet only ever learns of a mint by being told its address. There is no way to find one, and nothing that might follow from finding one, a list or a recommendation, can exist until that first step does.

With MONEYER_ANNOUNCE=true, and the same Nostr identity zap-to-note uses, the mint announces itself on the same hourly pass as the snapshot: kind 30078 again, d tag lnurlcash-mint. The content is this mint's own discovery document, exactly as /.well-known/lnurlw/<user> serves it, so there is one description of a mint rather than two that drift apart. It carries a sig over the canonicalised (RFC 8785) document made with the note signing key, so a holder can check that the mint announcing itself is the mint their notes verify against, whoever the Nostr identity publishing it belongs to. verifyAnnouncement in this package does that check.

Off by default, deliberately. A mint that does not want to be listed says nothing, and turning it on is the operator saying otherwise. Announcing needs MONEYER_PUBLIC_ORIGIN, since an announcement made with a Host header nobody sent is an address nobody can call back on.

There is no recommendation or review here, and no new event kind. Both are protocol decisions that belong with the LNURLcash NIPs rather than with a single mint's implementation.

The mint fee, and the rounding question

LUD-25 gives the fee as a flat base_fee_msat plus a ppm cut and says nothing about rounding. dni's lnurl-mint, the reference, ceilings it to a whole sat on purpose so the mint is never short a sat; moneyer withholds the msat-exact amount. Neither is out of spec - lnurlcash-kit's mintFeeBand treats anything between the two as the mint keeping its word, and the conformance grader accepts both.

moneyer rounds, by default. A mint that deals in fractions of a sat issues notes nothing downstream can spend: msat-exact fees make notes like 94.9 sat, and most Lightning wallets can only invoice whole sats, so such a note cannot be withdrawn by them at all. moneyer covers the notes already out there by advertising minWithdrawable as the note floored to a whole sat and accepting a melt for that, keeping the sub-sat remainder as dust - a wart it should not need. Rounding at mint time means it never comes up, and every note this mint issues for a whole-sat payment is worth a whole number of sats.

MONEYER_ROUND_FEE_TO_SAT=false restores the msat-exact fee. An operator upgrading into this default withholds slightly more per mint than before, which is why the changelog says so plainly rather than letting a redeploy change it quietly.

Melting to an invoice with no amount

An invoice that states no amount leaves the sending side to decide, and for a bearer note there is nothing to decide: the note's value is the amount. moneyer melts a note to one for the whole-sat floor of its value, the same figure it accepts from a wallet that fills the amount in itself, and keeps the sub-sat remainder as the same dust described above.

A note worth less than a single satoshi has nothing left once the floor applies, so an amountless invoice from one is refused with insufficient value. It can still be melted the ordinary way by a wallet that invoices its exact millisatoshi value.

Nothing else about the melt changes. OK still means in flight, the note still burns only on confirmed payment and restores only on confirmed non-payment, and the routing budget is still sized against the note.

Name the note you are buying

Paying a mint invoice mints a note, and by default that note's spend secret is the invoice's payment preimage. A payment preimage is not a private thing. The funding source has it, every node that forwarded the payment has it, and LUD-21 verify hands it to whoever asks with the payment hash, which is written inside the invoice on the payer's screen. The draft's answer is for the wallet to claim and rotate the instant the invoice settles, which is a foot race the wallet has to keep winning.

So a wallet may name the note instead. It chooses the secret first, writes it down, and sends h, the sha256 of that secret, on the pay callback:

GET /p/cb?amount=21000&h=<64 hex>

The mint credits the note at h when the invoice settles. The payment preimage then buys nothing: it is an ordinary payment proof, which is why verify goes on serving it. Nobody but the buyer ever knew the secret, so there is no race left to run, and no window in which holding the invoice is nearly holding the money.

h is optional and additive. A wallet that sends none gets exactly the behaviour it always got, so upgrading this mint breaks nothing that works today, and the LUD-25 draft needs no change to allow it.

The rules:

  • h is 64 hex characters, the sha256 of a 32-byte secret: the same meaning h carries on the withdraw callback. Upper case is accepted and read as lower case, there too.
  • A malformed h is refused before the mint asks its funding source for anything, so a wallet is never left holding a quote the mint was always going to reject.
  • An h that already names a note, an invoice, or a note another buyer has already bought is refused with Invalid or already spent k1., the same sentence a colliding output gets on the withdraw callback. Which table an id sits in is an oracle nobody is owed.
  • The callback's reply carries mintToHash: true when the invoice really was bound to h. A mint that did not implement the parameter would ignore it and answer without that field, so a wallet can tell the two apart before paying rather than by hunting for a note afterwards.
  • The payRequest and the discovery document both advertise mintToHash: true, so a wallet knows this mint takes the parameter before it asks rather than after it pays.
  • Claiming needs nothing else. GET /w?k1=<the secret> brings the note into existence as soon as the invoice has settled, with no verify poll and no preimage involved. The poll is still the way to claim from a mint that does not advertise mintToHash.
  • Persist the secret before asking for the invoice. Paying and then losing the secret is the one way this is worse than the old arrangement, and writing it down first removes it entirely.

A named note is also derived-secret friendly: a wallet whose secrets come from its seed can restore a note it bought but never claimed, which a note whose secret was a preimage could never offer.

A retried mutation is answered, not refused

Rotate, split and merge are GETs, and transports retry GETs. Go's net/http retries one that failed on a reused idle connection; the JDK's HttpClient retries idempotent methods with no switch to stop it. The retry is byte-identical, and it arrives after the mint has already burned the inputs.

Answering it "already spent" is what destroys money. A wallet that believes the refusal - which is exactly what the error taxonomy tells it to do - deletes the staged secret, and the mint keeps a note nobody can ever spend.

So moneyer records which request minted which outputs, and a request that has already minted its outputs gets the same reply again: same OK, same signatures, nothing burned, nothing minted, no balance moved. A retry is a read.

A request is the same request when it names the same input notes and asks for the same outputs: the same h, the same h2, the same amount. Input order does not matter, because a reordered retry is the same operation. Anything else naming a burned note is a double-spend attempt and still gets Invalid or already spent k1., unchanged. Provenance is recorded rather than inferred for that reason: matching on "a note exists at h" alone would let anyone holding a burned k1 and any outstanding note id draw a success out of the mint.

The melt path is untouched: melts are deduplicated by payment hash, which is a different question with a different answer.

This is moneyer's own behaviour. The LUD-25 draft says nothing about retries yet; the suggested wording is on lnurl/luds#301 as a SHOULD.

What the mint knows

This mint knows every note it has issued and what each is worth. It knows every rotate, split and merge, the links between them - which note became which, and when - and the network address the request came from. It knows the invoice a melt paid.

It does not know who holds a note between those operations. A note handed to someone else offline leaves no trace here until they rotate it, which is one reason a wallet rotates on receipt.

The wallet-side mitigations are weak, and worth naming as weak. A Tor or SOCKS proxy hides the address, not the links. Rotating at unpredictable times blurs the timing, not the graph. Nothing a holder does stops this mint seeing the chain of notes it struck.

The design was chosen anyway because it needs no new cryptography, any LUD-03 wallet can cash a note out, and verifying a note offline needs a signature and nothing else. The privacy story is trust the operator, and that is worth saying plainly.

Zap-to-note: a lightning address that pays out as a note

A Nostr zap is an ordinary LNURL-pay. Paid to the mint's own address it would mint a note, but to the payer: the invoice preimage is the secret and on Lightning the payer always learns it. So moneyer can also serve names that work the other way round. A zap to alice@<host> gets an invoice with a throwaway preimage; when it settles, the mint creates a note with a fresh secret, seals it in a NIP-59 gift wrap (a kind 2525 rumor, the shape heartwood-esp32 and notecase read) to alice's pubkey, leaves it on her NIP-17 inbox relays, and publishes the kind 9735 receipt so the zap shows up in her client like any other. A hardware wallet that catches up on its inbox when it powers on will find the note waiting.

Until alice rotates the note, the mint knows its secret. That is the position every freshly minted note is in, and it is why wallets rotate on receipt; notecase does it on heartwood collect. What is new is that the mint learns who was paid, which a lightning address always did.

MONEYER_PUBLIC_ORIGIN=https://mint.example      # required: settlement is on a timer
MONEYER_NOSTR_KEY=<32 bytes hex>                # the mint's own Nostr identity
MONEYER_NOSTR_RELAYS=wss://relay.example,wss://nos.lol
MONEYER_ZAP_NAMES="alice=npub1...,bob=<hex pubkey>"   # optional, see below

The key and the relays go together; the names are optional, because a mint can open registration and start with none of its own. A zap name's payRequest carries allowsNostr and the mint's nostrPubkey, and deliberately no withdrawLink. The receipt carries no preimage tag: it is optional in NIP-57, and here it would only invite someone to mistake it for the note. A zap name must not be the mint username or _. The kind 2525 rumor carries the zap request in a description tag, so a wallet can show who zapped and what they wrote without fetching the receipt.

Self-service: anyone with an npub can claim a name

Set MONEYER_NAME_PRICE_MSAT and the mint takes registrations. Unset means registration is closed and POST /names 404s; 0 means free, capped at three names per pubkey (rate limiting per address is the reverse proxy's job).

curl -X POST https://mint.example/names \
  -H "Authorization: Nostr <base64 kind 27235 event>" \
  -d '{"name":"donkey","note":"https://mint.example/w?k1=...&amount=21000"}'

The NIP-98 event must be kind 27235, signed within 60 seconds, with u set to the full URL, method set to POST, and payload set to the SHA-256 of the exact body sent. The key that signs it owns the name. No pubkey in the body is accepted, and there is no account, no password and no recovery.

A paid registration hands over a note of this mint, as a URL or as a bare secret. The mint burns the whole note - liabilities drop and the sats are revenue - so send one worth exactly the price; split first if you need to. A note worth less than the price is refused, and so is one from another host. Nothing is burned unless the name is granted: the burn and the registration are one transaction.

Names are ^[a-z0-9][a-z0-9_.-]{2,31}$. The mint's own username, _, admin and mint are reserved, as is any name already taken. Names are permanent in this version; removing one is the operator's, through moneyer admin names rm.

A registered name resolves on both rails at once:

  • [email protected] as a lightning address, paying out as a bearer note gift-wrapped to the owner's key.
  • GET /.well-known/nostr.json?name=<name> as a NIP-05 address. Only the name asked for is answered: the list of everyone here is not something to hand out.

The discovery endpoint advertises namePriceMsat while registration is open, so a wallet can offer the flow without asking.

Reaching the mint over Tor

Set MONEYER_ONION_URL to this mint's hidden service address and a request that arrives on that host gets every URL built from it: the callback, the withdrawLink, the payLink, the lightning address in the metadata.

Without it a Tor visitor is answered with the clearnet origin, and their wallet is then told to leave Tor to finish the job - the callback fetch goes out over clearnet, from their address, naming the mint they bank with. That is worse than not working, because it looks like it worked.

The Host header chooses between origins here; it never builds one. MONEYER_PUBLIC_ORIGIN exists precisely because the header is attacker controlled, and that property is unchanged: the worst a forged Host can do is get back the onion URL instead of the clearnet one, and the operator configured both. Anything else falls through to MONEYER_PUBLIC_ORIGIN.

Serving both audiences at once

Both doors serve the same notes. A note is keyed by sha256(k1) and not by host, so a k1 struck on the onion is valid at the clearnet host and the other way round. What is bound to a host is only the URL a note travels as - and an onion note handed to somebody without Tor is unspendable to them, while a clearnet note redeemed over Tor is private only if their wallet proxies.

So when both are configured, /w and the mint address name the other one in mirrors:

{"tag": "withdrawRequest", "callback": "http://<v3>.onion/w/cb",
 "mirrors": ["https://mint.example"], "mintPubkey": "..."}

Absent when there is only one door, rather than an empty array - a wallet reading [] would think it had been told something. The hourly Nostr announcement carries it too, so a wallet that discovers this mint on a relay learns both doors before it holds anything.

A wallet must not fail over to a mirror on its own. Every request that identifies a note carries the k1, so switching hosts discloses a bearer secret to the new one, and nothing in LUD-25 lets a host prove it is the same mint before receiving that secret - a hostile discovery document can name any host and claim any mintPubkey, because claiming is not proving. Show the holder the other door and let them choose it. Closing that properly needs a challenge the mint signs, which the spec does not have yet.

Running the hidden service itself is not this mint's job. Point a Tor HiddenServiceDir (or a proxy that speaks onion) at whatever host and port it already listens on, the same way you would front it with Caddy for clearnet.

Endpoints

| | | | --- | --- | | /.well-known/lnurlp/<user> | LUD-16 payRequest; paying mints a note | | /.well-known/lnurlp/<zap name> | NIP-57 payRequest; paying mints a note to the name's pubkey | | /.well-known/lnurlw/<user> | LUD-25 mint address discovery (experimental) | | /p/cb | LUD-06 pay callback; issues the mint invoice, and takes an optional h naming the note | | /z/cb/<zap name> | the zap callback; validates the kind 9734 and issues the invoice | | /verify/<hash> | LUD-21 verify, for mint invoices and melt payments | | /w | LUD-03 informational GET; a live note also carries payLink, the route back to this mint's discovery document | | /w/cb | the mutating callback: melt, rotate, split, merge | | POST /names | claim a lightning address, authenticated by NIP-98 | | /.well-known/nostr.json | NIP-05 for the names this mint serves | | /stats | what the mint owes, what the node holds, and the coverage between them | | /metrics | the same figures for a scraper, off by default |

Operating

set -a; source /etc/moneyer/env; set +a
moneyer admin status

moneyer admin reads the same MONEYER_* environment the mint does, so there is one description of a deployment and not two. It opens the database read-only unless the command mutates: an operator poking at a live mint should not be able to write to it by accident, and a mistyped path should not silently create an empty database to answer from.

| command | does | | --- | --- | | status | liabilities, melts in flight, unsettled invoices, node balance and coverage, lifetime totals, keys | | notes [--state outstanding\|pending\|burned] [--limit n] | list notes, newest first | | note <id\|k1> | one note; 64 hex that names no note is hashed and looked up as the secret | | melts [--pending] | list melts | | reconcile | one pending-melt reconcile pass, printing what changed | | sweep | delete mint invoices whose expiry is provably past | | snapshot <path> | a consistent copy of the database, taken live, refusing to overwrite | | names list\|add <name> <npub>\|rm <name> | the lightning addresses this mint pays out as notes | | keys rotate | generate a signing key and print the two environment lines; writes nothing | | verify-note <url> | check a note's signature offline, then say what the mint holds at that id |

snapshot uses SQLite's VACUUM INTO, so it needs no sqlite3 binary on the box and is safe to run against a live mint.

Metrics

MONEYER_METRICS=true serves GET /metrics in the OpenMetrics text format: moneyer_outstanding_msat, moneyer_outstanding_notes, moneyer_pending_melts, moneyer_oldest_pending_melt_seconds, moneyer_local_balance_msat, moneyer_unsettled_mint_invoices, moneyer_mints_total, moneyer_melts_total{outcome} and moneyer_zaps_total.

The app does not authenticate it. Restrict the path to your scraper at the reverse proxy, the way /stats is left public deliberately and this is not.

The three alerts worth having

  • oldest pending melt over 30 minutes. A melt that cannot be resolved is a note nobody can spend and a payment nobody can account for. The reconciler retries every five minutes; half an hour without a terminal answer means the funding source needs looking at.
  • coverage under 1. The node cannot pay out every note it owes. Whether that is a channel imbalance or something worse, it is the one number a custodial mint must never be relaxed about.
  • unsettled invoices growing. Invoices are issued and never paid all day, and the sweep clears the expired ones. A count that climbs through a sweep means invoices are being issued that nobody can pay.

Testing

npm test          # vitest: protocol, signing, melt discipline, conformance

The conformance grader runs inside the suite against an in-process mint, read-only and spending paths both. The cln and lnd backends are direct ports of the reference mint's semantics but have not yet been exercised against live nodes; treat them as beta until they have.

Dogfood

Part of the ForgeSworn LNURLcash stack: built on lnurlcash-kit (fee maths, hashing, signature agreement) and farrier-kit (BOLT-11 decoding and preimage verification on the melt path). The companion wallet is @forgesworn/notecase.

Other mints, wallets and libraries speaking the same protocol are indexed in awesome-lnurlcash.

Licence

MIT.