diff --git a/docs/COMPARISON-lnbits-nostr_bunker.md b/docs/COMPARISON-lnbits-nostr_bunker.md
new file mode 100644
index 0000000..5aa6d77
--- /dev/null
+++ b/docs/COMPARISON-lnbits-nostr_bunker.md
@@ -0,0 +1,108 @@
+# nsecbunkerd vs. `lnbits/nostr_bunker`
+
+A comparison of this daemon (the aiolabs fork of `kind-0/nsecbunkerd`) against the
+upstream LNbits extension [`lnbits/nostr_bunker`](https://github.com/lnbits/nostr_bunker).
+
+> Source verified 2026-06-19 against `lnbits/nostr_bunker@main` (`services.py`,
+> `models.py`, `crud.py`). The two projects share a name and a NIP (NIP-46 remote
+> signing) but are architecturally **inverted**: this daemon *uses* LNbits as a
+> downstream wallet provider; the upstream extension *is* an LNbits extension that
+> turns a wallet account into the bunker.
+
+## The one thing that matters: where the nsec lives
+
+| | nsecbunkerd (this fork) | `lnbits/nostr_bunker` |
+|---|---|---|
+| **Signing key location** | On the **daemon** host, separate process from LNbits | On the **LNbits** host, inside the extension DB |
+| **At-rest protection** | Passphrase-encrypted (LND-style unlock) for manually-added keys | **Plaintext** in `nostr_bunker.bunkers_data.nsec` — no encryption |
+| **Integration direction** | LNbits is a *downstream dependency* (wallet factory) | LNbits is the *host* (wallet account = signer identity) |
+
+`crud.py:create_bunkers_data()` writes the nsec straight through
+`db.insert("nostr_bunker.bunkers_data", ...)` with no encryption step; `models.py`
+`BunkersData.nsec` is "the normalized private key stored directly." This is the exact
+posture the aiolabs roadmap (`aiolabs/lnbits#18`, "no nsec at rest on LNbits") exists
+to eliminate: the LNbits host runs extension code, payment plumbing, and a public API,
+so disk/root compromise there must NOT equal Nostr-identity compromise. The
+standalone-daemon model keeps signing off that host; the upstream extension puts the
+key right back on it, unencrypted.
+
+## Full side-by-side
+
+| Dimension | nsecbunkerd (this fork) | `lnbits/nostr_bunker` |
+|---|---|---|
+| **Form factor** | Standalone Node daemon (own process/container) | LNbits extension, runs inside the LNbits process |
+| **Stack** | TypeScript + NDK 3.0.3 + nostr-tools 2.20 + Prisma/SQLite | Python + Vue/Quasar UMD frontend |
+| **Relay transport** | Daemon opens its own relay connections (NDK); per-key kind:24133 subs pinned to explicit relays (#21) | Piggybacks the `nostrclient` extension's shared relay layer (`nostr_client.relay_manager.publish_message()`) |
+| **Tenancy** | Multi-key, multi-domain, multi-user from one daemon | One bunker per wallet account; multiplexes clients via multiple `bunker://` URLs |
+| **Admin / control plane** | Whitelisted admin npubs over E2E-encrypted Nostr events; separate bunker key holds no user key material; optional remote `app.nsecbunker.com` UI | LNbits admin UI; wallet owner is implicitly the operator |
+| **Account provisioning** | OAuth-like flow: remote `create_account` → NIP-05 file write → NIP-89 (`kind:31990`) announce → mints LNbits wallet via `usermanager` API + nostdress `lud16` | None — the LNbits account already exists; the wallet *is* the identity |
+
+## NIP-46 surface
+
+Both implement NIP-46 over kind:24133 and accept **both** NIP-04 and NIP-44 v2
+(upstream `services.py` tries `nip44_decrypt` first, falls back to `nip04_decrypt`).
+
+| Method | nsecbunkerd | `lnbits/nostr_bunker` |
+|---|---|---|
+| `connect` | ✓ | ✓ (returns secret/ack after permission check) |
+| `get_public_key` | ✓ | ✓ |
+| `sign_event` | ✓ (ACL-gated, wire-name vocab #14) | ✓ (`_assert_method_allowed` + auto/confirm flow) |
+| `nip04_encrypt` / `decrypt` | ✓ | ✓ |
+| `nip44_encrypt` / `decrypt` | ✓ | ✓ |
+| `ping` | ✓ | ✓ (`pong`) |
+| `switch_relays` | — | ✓ (returns relay list as JSON) |
+
+## Policy / permission model
+
+This is where the designs genuinely diverge, and where upstream has something worth
+borrowing.
+
+**nsecbunkerd** — relational ACL across several tables:
+- `KeyUser` — a (keyName, userPubkey) grant
+- `SigningCondition` — per-method/kind/content allow rules
+- `Policy` / `PolicyRule` — reusable rule sets with per-rule `maxUsageCount` + expiry
+- `Token` — redeemable connection grant bound to a policy, with `redeemedAt` / `revokedAt`
+- Live-policy auth re-evaluated at request time (#11)
+
+**`lnbits/nostr_bunker`** — policy is **the `bunker://` URL itself**. Each `UrlData`
+row carries its own:
+- `relays`, `secret`, `client_pubkey`
+- `permissions` (e.g. `sign_event:{kind}`), `can_read`, `can_write`
+- `auto_sign` (default `False`) vs `confirm_sign` (default `True`)
+- `expires_at`
+- `post_rate_limit_per_day` — daily cap on kind:1, enforced by counting
+  `get_signing_requests_since()` over 24h (`_assert_post_rate_limit`)
+
+Pending approvals live in `SigningRequest` (status: pending/approved/signed/rejected/error),
+mirroring this fork's `Request` + manual-approval flow.
+
+**Takeaway:** upstream's "one bunker, many scoped URLs, each URL is a self-contained
+grant" is arguably cleaner than this fork's `Token`+`Policy`+`SigningCondition` triad
+for the common case of "issue a narrowly-scoped grant to one client." If the ACL surface
+here is ever simplified, that URL-as-grant model is the reference design — note in
+particular the built-in `post_rate_limit_per_day`, which this fork has no direct
+equivalent for.
+
+## Where each fits the aiolabs stack
+
+- **nsecbunkerd is the signer; LNbits is a client of it.** This is the `#18` endgame:
+  LNbits routes signing through a `RemoteBunkerSigner` over NIP-46 (the
+  protocol-over-loopback boundary chosen deliberately over a Unix socket), and every
+  nsec — operator *and* server identity — is retired from the LNbits host.
+
+- **`lnbits/nostr_bunker` is the convenience inversion we're explicitly avoiding.**
+  Useful prior art for per-URL policy ergonomics, but adopting it as the *signer
+  location* would reintroduce plaintext nsec-at-rest on the payments host — the precise
+  thing `#18` is designed to kill.
+
+## Gaps to track on our side
+
+1. **OAuth-created keys are stored recoverable, not encrypted.**
+   `create_account.ts` writes `currentConfig.keys[keyName] = { key: key.privateKey }`,
+   unlike the passphrase-encrypted path the SECURITY-MODEL doc describes for
+   manually-added keys. The doc promises non-exfiltratable keys; the OAuth path doesn't
+   meet that bar. (We're still strictly better than upstream, which stores *all* nsecs
+   plaintext — but the doc/behavior gap is real.)
+
+2. **No per-grant rate limiting.** Upstream's `post_rate_limit_per_day` is a clean
+   primitive we lack. Worth considering as a `PolicyRule` field.
diff --git a/docs/acl-prior-art-survey.md b/docs/acl-prior-art-survey.md
new file mode 100644
index 0000000..01a9f57
--- /dev/null
+++ b/docs/acl-prior-art-survey.md
@@ -0,0 +1,297 @@
+# ACL prior-art survey — NIP-46 bunker implementations
+
+Source-verified survey of how other open-source NIP-46 remote signers model
+authorization and grant lifecycle, run to inform the #25 ACL redesign (enforce
+token + grant lifecycle live at sign time instead of via a materialized cache).
+
+> **Verification status.** Every claim below was read against actual source on
+> 2026-06-19 (clones at the commits noted per project). An initial automated
+> survey overstated several implementations (notably "Signet enforces all
+> lifecycle live" — false); the corrections are called out inline. Treat the
+> file:line citations as the authority, not the prose summaries.
+
+## TL;DR for the redesign
+
+- **Amber** is the one *positive* live-lifecycle template: store the absolute
+  deadline on the grant row, recompute the verdict against `now()` on every
+  request, treat the periodic sweep as cleanup only. It also time-boxes
+  *denials*, not just grants.
+- **Signet** (a fork of our own codebase) re-shipped our #24 bug — proof that
+  materializing a policy photocopy without a live join cannot enforce
+  grant-level TTL/usage. Its schema is still the best reference for the
+  token/policy decomposition (minus the one `applyToken` materialization line).
+- **FROSTR** has the cleanest *revocation decomposition* (3 independent layers)
+  and a good auditable-credential table — but enforces **no** live expiry
+  anywhere.
+- **promenade** confirms the **revoke = re-key** anti-pattern to avoid, and
+  debunks "FROST can't decrypt DMs" (it's a design choice, not a math limit).
+- **NDK** (which we embed) is a deliberately *blank* permit seam: we own 100% of
+  policy — and `get_public_key` bypasses the seam entirely (see #26).
+
+Decision unchanged: **Option D, leaning D1.** Amber = live-evaluation reference;
+Signet = schema reference; FROSTR = revocation-decomposition reference; NDK =
+confirmed blank seam.
+
+## Strategic decision: keep our fork, treat Signet as a parts donor (2026-06-19)
+
+Signet is a fork/re-architecture of the same kind-0/nsecbunkerd lineage we
+maintain, and is feature-richer on the standalone-operator surface (trust dial,
+suspension, NIP-49 at-rest, two-tier tokens, kill-switch, React dashboard,
+Android companion). We considered adopting it wholesale. **Decision: no — keep
+our fork as what we ship; lift Signet's patterns as needed.**
+
+Why:
+- **Replacing doesn't solve #25.** Signet re-ships our exact #24 (materialized
+  photocopy, no live grant-level join). We'd still have to do the live-join work
+  — after paying a migration cost.
+- **We'd lose the integration that makes it ours.** LNbits wallet provisioning
+  (`usermanager` + nostdress), the OAuth-like `create_account` flow, and being
+  the signer target for the #18 `RemoteBunkerSigner` endgame. Porting those into
+  Signet just means maintaining a fork of a more opinionated upstream.
+- **Lineage/bus-factor.** Our `master` tracks the canonical kind-0 upstream;
+  Signet is a solo-maintainer rewrite with choices we may not want (removed JWT
+  auth, Android surface). For a security-load-bearing component that's more risk,
+  not less.
+
+Why it's low-stakes either way: LNbits ↔ bunker is **NIP-46 over the wire** (the
+deliberate protocol-over-IPC choice), so the signer is substitutable by design.
+If our fork ever becomes a maintenance burden we can drop in any conformant
+NIP-46 signer (Signet, Amber-as-bunker, HSM-backed) with config-only changes —
+**not a one-way door.**
+
+Escape hatch (option 3, parked): run Signet unmodified behind the protocol. Only
+attractive if the LNbits provisioning/OAuth flows move out of the bunker into
+LNbits proper (plausible under #18), which would shrink the integration gap
+that's the main reason to stay. Revisit if #25 implementation reveals our
+daemon's NDK/relay/ACL plumbing is materially rougher than Signet's.
+
+---
+
+## A — daemon/server implementations with a real policy model
+
+### Signet — `Letdown2491/signet` (TS daemon + React UI + Kotlin companion)
+MIT, very active (v1.11.0, 2026-06). An extensive re-architecture of the same
+kind-0/nsecbunkerd codebase we maintain.
+
+- **Re-ships our #24.** `applyToken` (`nip46-backend.ts:807`) checks
+  `Token.expiresAt` once at redeem (`:895`), then materializes `policy.rules`
+  into lifecycle-free `SigningCondition` rows (`:845-862`); the sign-time path
+  (`acl.ts:checkRequestPermission`) never reads `Token` again.
+  `maxUsageCount`/`currentUsageCount` are touched only in the policy CRUD route —
+  never enforced. Same materialization-drift bug as ours.
+- **What it adds over us:** a coarse-cache layer for **subject-level** state on
+  `KeyUser` — `revokedAt`, `suspendedAt`/`suspendUntil`, `trustLevel` — read live
+  per request and invalidated on change (`invalidateAclCache`). Genuinely fixes
+  live *revoke* (our sibling spirekeeper#22). Puts revoke on **`KeyUser`, not
+  `Token`** — corroborating our revoke=subject / expiry=grant split.
+- **Trust dial** over a kind-risk classifier: `trustLevel ∈ {paranoid,
+  reasonable, full}`, `SAFE_KINDS` auto / `SENSITIVE_KINDS` (0/3/4/5/wallet/
+  auth/NIP-04) forced manual (`acl.ts:129-161`).
+- **Two-tier tokens:** one-time `ConnectionToken` (mandatory `expiresAt`,
+  validates connect but never auto-approves) vs policy-backed `Token` (atomic
+  claim `updateMany where redeemedAt:null`, `nip46-backend.ts:813`).
+- **Key-at-rest:** NIP-49 ncryptsec + AES-256-GCM envelope (PBKDF2-SHA256 @600k).
+- **Takeaway:** adopt its `KeyUser` subject-state + `Request` indexing; reject
+  its `applyToken` materialization; the `ConnectionToken`-vs-`Token` split *is*
+  D1 in schema form.
+
+### Amber — `greenart7c3/Amber` (Android, Kotlin/Room) ⭐ live-lifecycle reference
+MIT, very active (last commit 2026-06-19). Android signer (NIP-55 intents **and**
+NIP-46 over relays). Listed in tier A despite being mobile because its permission
+model is the strongest of any surveyed.
+
+- **Grant schema** (`ApplicationPermissionsEntity.kt:18-41`): unique composite
+  index over `(pkKey, type, kind, relay)` — per-(app × method × kind × relay).
+  Columns include `acceptable: Boolean`, `rememberType: Int`, `acceptUntil:
+  Long`, `rejectUntil: Long`.
+- **Expiry enforced LIVE** (the key finding): `IntentUtils.isRemembered()`
+  (`IntentUtils.kt:1087-1101`) is the per-request verdict and recomputes
+  `acceptUntil > TimeUtils.now()` / `rejectUntil > now()` fresh every call;
+  expired → returns `null` → falls through to a user prompt. Called on both the
+  NIP-46 relay path (`EventNotificationConsumer.kt:440-441`) and the NIP-55
+  intent path (`SignerProviderQuery.kt:183` etc.).
+- **The sweep is non-load-bearing.** `updateExpiredPermissions(time)`
+  (`ApplicationDao.kt:51`, exempts `rememberType <> 4`=ALWAYS) runs every 24h via
+  WorkManager — pure cleanup; correctness doesn't depend on it firing because the
+  decision is recomputed against `now()` on read.
+- **Time-boxed denials too:** `rejectUntil` means "reject for 5 min" decays back
+  to a prompt rather than a permanent no — a nicer primitive than a single
+  allow/deny flag.
+- **Wildcard-as-distinct-tier:** lookup ladder is exact-kind → all-kinds
+  (`kind IS NULL`, `getPermissionAllKinds`, `ApplicationDao.kt:87-91`); relay
+  wildcard matches `'*' OR '' OR NULL` in one query (`getWildcardRelayPermission`,
+  `:101-106`). Wildcard rows are explicitly queried, never an accidental
+  missing-WHERE match.
+- **Read-through LRU caches rows, not verdicts** (`CachingApplicationDao`) — keeps
+  the live `now()` re-check on every cache hit; invalidation is write-driven and
+  coarse per-app.
+- **Sign policies** (`ChooseSignPolicy.kt:32-45`, stored as `signPolicy: Int`):
+  `0` basic / `1` manual-per-new-app / `2` fully-auto (short-circuits to allow
+  before any row lookup, `IntentUtils.kt:1090`).
+- **Key-at-rest** (`SecureCryptoHelper.kt`): Android Keystore AES-256-GCM, 96-bit
+  IV / 128-bit tag, StrongBox-backed when available with TEE fallback and a
+  MediaTek denylist; optional app-level biometric gate.
+- **NIP-46 coverage** (`SignerType.kt`, `BunkerRequestUtils.kt:232-248`): connect,
+  sign_event, nip04/nip44 (+v3) encrypt/decrypt, get_public_key,
+  decrypt_zap_event, ping, switch_relays, sign_psbt, logout; both `bunker://` and
+  `nostrconnect://`.
+- **Steal for us:** absolute-deadline-on-row + recompute-vs-now per request;
+  time-boxed denials; wildcard as a distinct explicitly-queried tier; cache rows
+  not answers.
+
+### FROSTR — `FROSTR-ORG/igloo-server` + `bifrost` (TS, FROST k-of-n)
+MIT. igloo-server v1.2.0 (2026-05-28); bifrost v2.0.2 (2026-01-24). Threshold
+Schnorr over Nostr; igloo-server exposes the NIP-46 endpoint, bifrost is the node
+SDK.
+
+- **Three independent authorization layers** (the prize):
+  1. **App NIP-46 policy** — `Nip46Policy { methods?, kinds? }` (`db/nip46.ts:8-11`),
+     sessions keyed `(user_id, client_pubkey)` (`:92`), checked live per request
+     (`service.ts:508-509, 766-795`). No TTL/expiry. Session revoke is **explicit**
+     (`status='revoked'`, `:792-826`); per-method/kind revoke is **implicit** (flip
+     boolean false, audited at `:722-790`).
+  2. **Peer-transport policy** — per-peer directional `allowSend`/`allowReceive`
+     (`util/peer-policy.ts:3-9`, `docs/PEER_POLICIES.md`), enforced in bifrost
+     `_filter`/`get_recv_pubkeys` (`client.ts:226-245`). **Correction:** it's
+     *default-allow + explicit per-peer deny + last-layer-wins*, not "deny-override".
+  3. **Operator API auth** — keys stored SHA-256 hash+prefix with `revoked_at`
+     (checked first, timing-safe) + `last_used_at/ip` (`migrations/..._api_keys.sql`,
+     `database.ts:815-1047`); Argon2id password hashing (`config/crypto.ts:26-31`).
+- **No layer enforces live expiry.** `nip46_requests.expires_at` exists but is
+  never populated; the only time-based enforcement is the in-memory derived-key
+  vault (TTL + bounded reads + zeroize, `auth.ts:359-459`).
+- **Key-at-rest:** DB mode AES-256-GCM in SQLite, PBKDF2-HMAC-SHA256 **@600k**
+  (corrected from "~200k", `config/crypto.ts:7-11`); headless mode = plaintext env
+  (`GROUP_CRED`/`SHARE_CRED`).
+- **Distributed veto** is real at the participation level (a co-signer withholding
+  its partial below threshold blocks the sig) but the default signer auto-signs
+  (`middleware: {}`, `client.ts:55`) — realizing a veto needs a custom
+  `middleware.sign` not shipped by default.
+- **Share rotation** (recover → re-split, same group npub, old shares can't
+  combine) exists as a **bifrost SDK primitive** (`generate_dealer_package`), **not**
+  as an igloo-server endpoint; recovery reconstructs the full nsec in memory and
+  `/api/recover` even returns it over HTTP (`routes/recovery.ts:147-157`).
+- **Steal for us:** the 3-layer revocation decomposition; audit-event-on-grant-
+  change; `revoked_at`-checked-first + last-used credential table.
+
+### promenade — fiatjaf (Go, FROST coordinator + signer split)
+Off GitHub; cloned from fiatjaf's nostr-git (`relay.ngit.dev/npub180c…/promenade.git`),
+HEAD `70ff8439` 2026-06-18. NIP-46 method logic lives in the pinned dep
+`fiatjaf.com/nostr` (`nip46.DynamicSigner`).
+
+- **Architecture:** khatru coordinator-relay doubles as the NIP-46 endpoint, runs
+  the FROST ceremony, holds a transport/handler key but **no shard**
+  (`account_registration.go:44` carries only `frost.PublicKeyShard`); separate
+  signer daemons each hold one shard; m-of-n with m≤20 (`:79`). Signing-ceremony
+  kinds 26430–26434; account registration is kind **16430** (replaceable).
+- **No encrypted DMs — by choice, not by math.** `DynamicSigner` recognizes
+  `nip44_encrypt`/`nip44_decrypt`/`switch_relays` (`dynamic-signer.go`), but
+  promenade hardwires `AuthorizeEncryption → false` (`coordinator/nip46.go:167`)
+  and `GroupContext.Encrypt/Decrypt → "not implemented"` (`sign.go:288-302`).
+  README: *"destroyer of encryption."* **Correction:** threshold ECDH is NOT
+  impossible for FROST — `frost/ecdh.go` implements `CreateECDHShare` /
+  `AggregateECDHShards`; it's simply not plumbed in.
+- **ACL:** `AuthorizeSigning` per sign_event (`coordinator/nip46.go:86`); named
+  profiles `["profile", name, secret, restrictions]` where restrictions is a
+  `nostr.Filter` but only `Kinds` + `Until` are enforced (`:139-159`). The secret
+  is a reusable bearer capability.
+- **Lifecycle:** per-profile `Until` is the only time-bound; **no revoke API** —
+  dropping one capability means re-publishing the whole kind:16430 account signed
+  by the **master nsec**. The **revoke = re-key anti-pattern** to avoid.
+- **Key-at-rest:** nsec sharded client-side (never whole), but shards stored
+  **plaintext** in each signer's BoltDB (`acceptor.go:209`); coordinator/signer
+  identity keys from plaintext env.
+- **Relevance:** confirms (1) keep grant-revoke independent of key rotation, and
+  (2) for the #18 "bunker for everything" endgame, threshold-protecting the server
+  identity wouldn't *mathematically* preclude DM decryption — but keeping ECDH on a
+  separate non-threshold key is the cheaper path.
+
+---
+
+## B — library/SDK signer seams
+
+### NDK — `nostr-dev-kit/ndk` (we embed this) @ `4b86acd` (2026-04-05)
+nip46 under `core/src/signers/nip46/`.
+
+- Backend `NDKNip46Backend` (`backend/index.ts:58`), client `NDKNip46Signer`
+  (`index.ts:60`).
+- Permit seam: `Nip46PermitCallback = (params: {id, pubkey, method, params?}) =>
+  Promise<boolean>` (`backend/index.ts:29-43`), invoked via overridable
+  `pubkeyAllowed()` (`:229-231`) from each strategy.
+- **`get_public_key` bypasses the seam** — `backend/get-public-key.ts:3-11`
+  returns the pubkey with no `pubkeyAllowed` call. (rust-nostr's `approve()` wraps
+  every method including this one.) See #26.
+- Signature verified before dispatch (`index.ts:181`); strategies swappable
+  (`setStrategy`, `:156-158`).
+- `applyToken(pubkey, token)` default-throws (`:166-168`), invoked by the connect
+  handler when a token is present (`connect.ts:21-24`) — token policy is the
+  embedder's job.
+- **No** built-in scoping/kinds/rate-limit/expiry/persistence — all policy lives
+  behind the one callback. We own 100% of the policy engine.
+
+### rust-nostr / nostr-sdk @ `e47b572` (v0.45.0-alpha.1)
+- `NostrConnectRemoteSigner` (`signer.rs:39`) + `NostrConnect` client.
+- Trait `NostrConnectSignerActions::approve(&self, public_key, req) -> bool`
+  (`signer.rs:342-345`), synchronous bool, wraps the **entire** request match in
+  `serve()` (`:201-202`) — gates every method **including** `get_public_key`.
+- FFI (uniffi/wasm) exposes **only the client** `NostrConnect`, not the backend —
+  no non-Rust embedding of the signer side.
+
+### nak — `fiatjaf/nak` bunker subcommand @ `483bf94`
+- Allow-list of client pubkeys (`BunkerConfig.Clients`), `--persist`s 0600 JSON.
+- Once authorized, **signs everything** — no method/kind scoping, no expiry, no
+  rate limiting. Notably its underlying lib computes a `harmless` (connect/
+  get_public_key/ping) vs dangerous (sign/encrypt/decrypt) hint that nak
+  **discards**. A bare always-sign baseline.
+
+---
+
+## C — clients / extensions (less relevant; novel UX only)
+
+- **keys.band** — Svelte Chrome extension (NIP-07): the one browser signer with
+  *time-bounded* authorization grants (allow-for-N-minutes/session). Relevant to a
+  TTL-grant UX.
+- **nos2x / nos2x-fox** (fiatjaf) — origin of the per-origin "remember / allow
+  this site" NIP-07 model; key stored ~plaintext in extension storage.
+- **Gossip** (Rust desktop) — not a bunker, but best-in-class key-at-rest:
+  passphrase-encrypted on disk, startup unlock, memory zeroed before free. Clean
+  `LocalSigner` envelope reference.
+- **Primal**, **nowser** (Flutter) — clients that also serve NIP-46/NIP-55; use the
+  standard `optional_requested_perms` per-method/per-kind grammar.
+
+---
+
+## D — not bunkers / dead
+
+- **`Letdown2491/nip46-relay`** — a NIP-46 *transport relay* (forwards opaque
+  blobs), no signing/authz. Appears next to Signet; easy to mistake for a signer.
+- **Keychat** — Signal-over-Nostr chat app; signs only its own events.
+- **python-nostr** — abandoned 2022, no NIP-46. (No Python library offers a
+  signer-side permission abstraction; a Python bunker means hand-rolling the
+  kind-24133 loop or driving rust-nostr via FFI — and the FFI exposes only the
+  client.)
+
+---
+
+## Patterns worth stealing — consolidated
+
+1. **Live evaluation (Amber):** absolute deadline on the grant row; verdict is a
+   pure function recomputed vs `now()` per request; sweep is cleanup-only. This is
+   Option D, proven in production.
+2. **Time-box denials too (Amber `rejectUntil`):** a deny decays to a prompt.
+3. **Wildcard as a distinct, explicitly-queried tier (Amber):** never a fuzzy
+   missing-WHERE match in the auto-decide path.
+4. **Cache rows, never verdicts (Amber `CachingApplicationDao`, Signet coarse
+   cache):** keep the `now()` re-check on every hit; invalidate on write.
+5. **Subject vs grant separation (Signet):** revoke/suspend/trust on `KeyUser`
+   (cheap, cache+invalidate); expiry/usage on `Token`/`Policy` (must join live).
+6. **Usage = COUNT(Request) in window (lnbits/nostr_bunker), not a mutable
+   counter:** drop `currentUsageCount`; needs `Request.keyUserId` + index.
+7. **Revocation decomposition (FROSTR):** app-grant revoke ≠ transport quarantine ≠
+   key rotation. Never collapse grant-revoke into re-key (promenade anti-pattern).
+8. **Auditable, revocable credentials (FROSTR):** `revoked_at` checked first +
+   last-used tracking; audit-event-on-grant-change decoupled from enforcement.
+9. **Single predicate `grantIsLive(now)`** used at both redeem and sign time
+   (the discipline that prevents the original drift).
+10. **NDK seam reality:** we own all policy; design around `get_public_key`
+    bypassing `pubkeyAllowed`.
diff --git a/package.json b/package.json
index 4975b68..f1f8732 100644
--- a/package.json
+++ b/package.json
@@ -21,6 +21,7 @@
   "scripts": {
     "build": "tsup src/index.ts; tsup src/daemon/index.ts -d dist/daemon; tsup src/client.ts -d dist/client",
     "build:client": "tsup src/client.ts -d dist/client",
+    "test": "TS_NODE_TRANSPILE_ONLY=1 node -r ts-node/register --test tests/*.test.ts",
     "prisma:generate": "npx prisma generate",
     "prisma:migrate": "npx prisma migrate deploy",
     "prisma:create": "npx prisma db push --preview-feature",
diff --git a/prisma/migrations/20260619125847_live_grant_lifecycle_schema/migration.sql b/prisma/migrations/20260619125847_live_grant_lifecycle_schema/migration.sql
new file mode 100644
index 0000000..7608e6e
--- /dev/null
+++ b/prisma/migrations/20260619125847_live_grant_lifecycle_schema/migration.sql
@@ -0,0 +1,37 @@
+-- RedefineTables
+PRAGMA defer_foreign_keys=ON;
+PRAGMA foreign_keys=OFF;
+CREATE TABLE "new_Request" (
+    "id" TEXT NOT NULL PRIMARY KEY,
+    "keyName" TEXT,
+    "createdAt" DATETIME NOT NULL DEFAULT CURRENT_TIMESTAMP,
+    "requestId" TEXT NOT NULL,
+    "remotePubkey" TEXT NOT NULL,
+    "method" TEXT NOT NULL,
+    "params" TEXT,
+    "allowed" BOOLEAN,
+    "keyUserId" INTEGER,
+    CONSTRAINT "Request_keyUserId_fkey" FOREIGN KEY ("keyUserId") REFERENCES "KeyUser" ("id") ON DELETE SET NULL ON UPDATE CASCADE
+);
+INSERT INTO "new_Request" ("allowed", "createdAt", "id", "keyName", "method", "params", "remotePubkey", "requestId") SELECT "allowed", "createdAt", "id", "keyName", "method", "params", "remotePubkey", "requestId" FROM "Request";
+DROP TABLE "Request";
+ALTER TABLE "new_Request" RENAME TO "Request";
+CREATE INDEX "Request_keyUserId_method_idx" ON "Request"("keyUserId", "method");
+CREATE TABLE "new_SigningCondition" (
+    "id" INTEGER NOT NULL PRIMARY KEY AUTOINCREMENT,
+    "method" TEXT,
+    "kind" TEXT,
+    "content" TEXT,
+    "keyUserKeyName" TEXT,
+    "allowed" BOOLEAN,
+    "keyUserId" INTEGER,
+    "createdAt" DATETIME NOT NULL DEFAULT CURRENT_TIMESTAMP,
+    "expiresAt" DATETIME,
+    "revokedAt" DATETIME,
+    CONSTRAINT "SigningCondition_keyUserId_fkey" FOREIGN KEY ("keyUserId") REFERENCES "KeyUser" ("id") ON DELETE SET NULL ON UPDATE CASCADE
+);
+INSERT INTO "new_SigningCondition" ("allowed", "content", "id", "keyUserId", "keyUserKeyName", "kind", "method") SELECT "allowed", "content", "id", "keyUserId", "keyUserKeyName", "kind", "method" FROM "SigningCondition";
+DROP TABLE "SigningCondition";
+ALTER TABLE "new_SigningCondition" RENAME TO "SigningCondition";
+PRAGMA foreign_keys=ON;
+PRAGMA defer_foreign_keys=OFF;
diff --git a/prisma/schema.prisma b/prisma/schema.prisma
index f0072ea..73f54df 100644
--- a/prisma/schema.prisma
+++ b/prisma/schema.prisma
@@ -17,6 +17,14 @@ model Request {
     method        String
     params        String?
     allowed       Boolean?
+    // Bind each request to the KeyUser it was evaluated against so usage
+    // caps can be derived live by COUNTing allowed Requests, instead of
+    // maintaining a mutable PolicyRule.currentUsageCount that drifts.
+    // See aiolabs/nsecbunkerd#25 (Option D, derive-don't-count).
+    keyUserId     Int?
+    KeyUser       KeyUser? @relation(fields: [keyUserId], references: [id])
+
+    @@index([keyUserId, method])
 }
 
 model KeyUser {
@@ -31,6 +39,7 @@ model KeyUser {
   logs              Log[]
   signingConditions SigningCondition[]
   Token             Token[]
+  requests          Request[]
 
   @@unique([keyName, userPubkey], name: "unique_key_user")
 }
@@ -56,15 +65,25 @@ model User {
   pubkey    String
 }
 
+// The SigningCondition layer is the MANUAL-OVERRIDE source of truth
+// (web-approval / add_signing_condition / create_account bootstrap) — it is
+// no longer materialized from token policies (see aiolabs/nsecbunkerd#25:
+// applyToken stopped photocopying; token grants are evaluated live off
+// Token -> Policy -> PolicyRule). Under D1 the override layer carries its
+// own lifecycle so it runs through the same grantIsLive(now) predicate as
+// token grants.
 model SigningCondition {
-  id             Int      @id @default(autoincrement())
+  id             Int       @id @default(autoincrement())
   method         String?
   kind           String?
   content        String?
   keyUserKeyName String?
   allowed        Boolean?
   keyUserId      Int?
-  KeyUser        KeyUser? @relation(fields: [keyUserId], references: [id])
+  createdAt      DateTime  @default(now())
+  expiresAt      DateTime?
+  revokedAt      DateTime?
+  KeyUser        KeyUser?  @relation(fields: [keyUserId], references: [id])
 }
 
 model Log {
diff --git a/src/daemon/backend/index.ts b/src/daemon/backend/index.ts
index 5d05a8a..91f2f58 100644
--- a/src/daemon/backend/index.ts
+++ b/src/daemon/backend/index.ts
@@ -1,6 +1,7 @@
 import NDK, { NDKNip46Backend, NDKPrivateKeySigner, Nip46PermitCallback } from '@nostr-dev-kit/ndk';
 import prisma from '../../db.js';
 import type {FastifyInstance} from "fastify";
+import { grantIsLive } from '../lib/acl/index.js';
 
 export class Backend extends NDKNip46Backend {
     public baseUrl?: string;
@@ -91,7 +92,10 @@ export class Backend extends NDKNip46Backend {
         if (!tokenRecord) throw new Error("Token not found");
         if (tokenRecord.redeemedAt) throw new Error("Token already redeemed");
         if (!tokenRecord.policy) throw new Error("Policy not found");
-        if (tokenRecord.expiresAt && tokenRecord.expiresAt < new Date()) throw new Error("Token expired");
+        // Revoke + expiry via the single grantIsLive predicate — the exact
+        // check the sign-time ACL uses, so redeem-time and sign-time cannot
+        // drift (the root of #24). See aiolabs/nsecbunkerd#25.
+        if (!grantIsLive(tokenRecord)) throw new Error("Token expired or revoked");
 
         return tokenRecord;
     }
@@ -100,39 +104,20 @@ export class Backend extends NDKNip46Backend {
         const tokenRecord = await this.validateToken(token);
         const keyName = tokenRecord.keyName;
 
-        // Upsert the KeyUser with the given remotePubkey
+        // Record ONLY the binding (KeyUser <- Token). Under #25 the token's
+        // policy is evaluated live at sign time (checkIfPubkeyAllowed step 4)
+        // off Token -> Policy -> PolicyRule, NOT photocopied into
+        // SigningCondition rows here. That photocopy was the root of #24: the
+        // copy carried no expiry/revoke and short-circuited the live check, so
+        // an expired or revoked token kept signing forever. With no copy, the
+        // token's lifecycle is re-checked on every request and there is nothing
+        // to keep in sync.
         const upsertedUser = await prisma.keyUser.upsert({
             where: { unique_key_user: { keyName, userPubkey } },
             update: { },
             create: { keyName, userPubkey, description: tokenRecord.clientName },
         });
 
-        await prisma.signingCondition.create({
-            data: {
-                keyUserId: upsertedUser.id,
-                method: 'connect',
-                allowed: true,
-            }
-        });
-
-        // Go through the rules of this policy and apply them to the user
-        for (const rule of tokenRecord!.policy!.rules) {
-            const signingConditionQuery: any = { method: rule.method };
-
-            if (rule && rule.kind) {
-                signingConditionQuery.kind = rule.kind.toString();
-            }
-
-            await prisma.signingCondition.create({
-                data: {
-                    keyUserId: upsertedUser.id,
-                    method: rule.method,
-                    allowed: true,
-                    ...signingConditionQuery,
-                }
-            });
-        }
-
         await prisma.token.update({
             where: { id: tokenRecord.id },
             data: {
diff --git a/src/daemon/lib/acl/index.ts b/src/daemon/lib/acl/index.ts
index f621c98..1c7912d 100644
--- a/src/daemon/lib/acl/index.ts
+++ b/src/daemon/lib/acl/index.ts
@@ -1,31 +1,44 @@
-import { NDKEvent, NostrEvent, NIP46Method } from '@nostr-dev-kit/ndk';
+import { NostrEvent, NIP46Method } from '@nostr-dev-kit/ndk';
 import prisma from '../../../db.js';
+import { liveWhere } from './lifecycle.js';
+
+// Re-export the single lifecycle predicate so callers (e.g.
+// Backend.validateToken) import it from the ACL module. The implementation
+// lives in ./lifecycle.ts so it can be unit-tested without a database.
+export { grantIsLive } from './lifecycle.js';
 
 /**
- * Layered authorization check. Order matters:
+ * Layered authorization check. Order matters (denials beat grants):
  *
  *   1. fetch KeyUser; if missing → undefined (no binding exists)
- *   2. if KeyUser.revokedAt set → false (binary user revoke beats everything)
- *   3. SigningCondition override layer (per-user grants/denies):
- *        - explicit reject (method='*', allowed=false) → false
- *        - matching per-(method,kind) row → return row.allowed
- *   4. Live policy join over KeyUser → Token → Policy → PolicyRule
- *        with Token.revokedAt IS NULL and a matching rule → true
- *   5. else → undefined (denied)
+ *   2. KeyUser.revokedAt set → false (subject-level ban beats everything)
+ *   3. manual-override layer (LIVE SigningConditions only):
+ *        - live matching per-(method,kind) deny → false
+ *        - live matching per-(method,kind) grant → true
+ *   4. live token grant: a redeemed Token bound to this KeyUser that is
+ *      neither revoked nor expired pairs the user (`connect`) outright and,
+ *      via its policy, governs signing. Token expiry/revoke are evaluated
+ *      HERE, every request — not photocopied at redeem (#24).
+ *   5. else → undefined (caller's requestPermission flow may prompt an admin)
  *
- * Step 3 must precede step 4: per-user denies override the policy, and
- * per-user grants extend beyond the policy. Step 2 must precede step 3:
- * a revoked KeyUser stays revoked regardless of conditions or policy.
+ * Unlike the pre-#25 algorithm, token grants are no longer materialized into
+ * SigningCondition rows at redeem (Backend.applyToken stopped photocopying),
+ * so step 4 is the live source of truth for token lifecycle. The override
+ * layer (step 3) is manual-only and now carries its own lifecycle, so an
+ * expired/revoked override stops granting too.
  *
- * See aiolabs/nsecbunkerd#11 and the issue comment that ratified the
- * algorithm (https://git.atitlan.io/aiolabs/nsecbunkerd/issues/11#issuecomment-1473).
+ * Supersedes the #11 algorithm; closes the materialization-drift family
+ * behind #24. See aiolabs/nsecbunkerd#25.
  */
 export async function checkIfPubkeyAllowed(
     keyName: string,
     remotePubkey: string,
     method: IMethod,
-    payload?: string | NostrEvent
+    payload?: string | NostrEvent,
 ): Promise<boolean | undefined> {
+    // One clock reading for the whole decision.
+    const now = new Date();
+
     // Step 1: find KeyUser.
     const keyUser = await prisma.keyUser.findUnique({
         where: { unique_key_user: { keyName, userPubkey: remotePubkey } },
@@ -35,81 +48,89 @@ export async function checkIfPubkeyAllowed(
         return undefined;
     }
 
-    // Step 2: binary user revoke.
+    // Step 2: subject-level revoke (sticky ban, beats everything).
     if (keyUser.revokedAt) {
         return false;
     }
 
-    // Step 3a: explicit-reject override (rejectAllRequestsFromKey writes this).
-    const explicitReject = await prisma.signingCondition.findFirst({
-        where: {
-            keyUserId: keyUser.id,
-            method: '*',
-            allowed: false,
-        }
+    const live = liveWhere(now);
+
+    // Step 3: live matching per-(method, kind) override — deny beats grant.
+    // (Subject-level "reject all from this user" is KeyUser.revokedAt, applied
+    // at step 2 via the revoke_user admin command. There is no method='*'
+    // SigningCondition sentinel — nothing writes one.)
+    const signingConditionQuery = requestToSigningConditionQuery(method, payload);
+
+    const liveDeny = await prisma.signingCondition.findFirst({
+        where: { keyUserId: keyUser.id, ...signingConditionQuery, allowed: false, ...live },
     });
 
-    if (explicitReject) {
-        console.log(`explicit reject`, explicitReject);
+    if (liveDeny) {
         return false;
     }
 
-    // Step 3b: matching per-(method, kind) override.
-    const signingConditionQuery = requestToSigningConditionQuery(method, payload);
-
-    const signingCondition = await prisma.signingCondition.findFirst({
-        where: {
-            keyUserId: keyUser.id,
-            ...signingConditionQuery,
-        }
+    const liveGrant = await prisma.signingCondition.findFirst({
+        where: { keyUserId: keyUser.id, ...signingConditionQuery, allowed: true, ...live },
     });
 
-    if (signingCondition && (signingCondition.allowed === true || signingCondition.allowed === false)) {
-        console.log(`found signing condition`, signingCondition);
-        return signingCondition.allowed;
-    }
-
-    // Step 4: live policy join. Walk every non-revoked Token bound to this
-    // KeyUser; if any of their policies has a matching PolicyRule, allow.
-    //
-    // PolicyRule.kind matching:
-    //   - exact match against payload kind (stringified — matches the
-    //     create_new_policy.ts:23 storage format `rule.kind.toString()`)
-    //   - 'all' literal matches any kind (parity with the override-layer
-    //     allowScopeToSigningConditionQuery convention)
-    //   - NULL kind is a defensive branch — no current code path inserts
-    //     PolicyRules with null kind, but if one ever appears (raw SQL,
-    //     future code, schema migration) we treat it as a wildcard rather
-    //     than failing closed silently.
-    const payloadKindString = (method === 'sign_event' && typeof payload === 'object' && payload?.kind !== undefined)
-        ? payload.kind.toString()
-        : undefined;
-
-    const kindMatchers: Array<{ kind: string | null }> = [{ kind: null }, { kind: 'all' }];
-    if (payloadKindString !== undefined) {
-        kindMatchers.push({ kind: payloadKindString });
-    }
-
-    const policyAllowance = await prisma.token.findFirst({
-        where: {
-            keyUserId: keyUser.id,
-            revokedAt: null,
-            policy: {
-                rules: {
-                    some: {
-                        method,
-                        OR: kindMatchers,
-                    },
-                },
-            },
-        },
-    });
-
-    if (policyAllowance) {
+    if (liveGrant) {
         return true;
     }
 
-    // Step 5: no override granted, no policy rule matched. Caller's
+    // Step 4: live token grant.
+    //
+    // A redeemed token that is live (not revoked, not past expiry) grants
+    // `connect` (the pairing) outright, and grants other methods when its
+    // policy has a matching PolicyRule. The live filter is what closes #24:
+    // an expired or revoked token simply stops matching here, every request,
+    // with no photocopy to outlive it.
+    if (method === 'connect') {
+        const liveToken = await prisma.token.findFirst({
+            where: { keyUserId: keyUser.id, ...live },
+        });
+
+        if (liveToken) {
+            return true;
+        }
+    } else {
+        // PolicyRule.kind matching:
+        //   - exact match against the stringified payload kind (matches the
+        //     create_new_policy.ts storage format `rule.kind.toString()`)
+        //   - 'all' literal matches any kind
+        //   - NULL kind is a defensive wildcard — no current writer emits a
+        //     null-kind rule, but treat it as a wildcard rather than failing
+        //     closed silently if one ever appears (raw SQL, future code).
+        const payloadKindString =
+            method === 'sign_event' && typeof payload === 'object' && payload?.kind !== undefined
+                ? payload.kind.toString()
+                : undefined;
+
+        const kindMatchers: Array<{ kind: string | null }> = [{ kind: null }, { kind: 'all' }];
+        if (payloadKindString !== undefined) {
+            kindMatchers.push({ kind: payloadKindString });
+        }
+
+        const policyAllowance = await prisma.token.findFirst({
+            where: {
+                keyUserId: keyUser.id,
+                ...live,
+                policy: {
+                    rules: {
+                        some: {
+                            method,
+                            OR: kindMatchers,
+                        },
+                    },
+                },
+            },
+        });
+
+        if (policyAllowance) {
+            return true;
+        }
+    }
+
+    // Step 5: no live override and no live token grant matched. Caller's
     // requestPermission flow may still prompt the admin out-of-band.
     return undefined;
 }
@@ -195,20 +216,3 @@ export async function allowAllRequestsFromKey(
         console.log('allowAllRequestsFromKey', e);
     }
 }
-
-export async function rejectAllRequestsFromKey(remotePubkey: string, keyName: string): Promise<void> {
-    // Upsert the KeyUser with the given remotePubkey
-    const upsertedUser = await prisma.keyUser.upsert({
-        where: { unique_key_user: { keyName, userPubkey: remotePubkey } },
-        update: { },
-        create: { keyName, userPubkey: remotePubkey },
-    });
-
-    // Create a new SigningCondition for the given KeyUser and set allowed to false
-    await prisma.signingCondition.create({
-        data: {
-            allowed: false,
-            keyUserId: upsertedUser.id,
-        },
-    });
-}
\ No newline at end of file
diff --git a/src/daemon/lib/acl/lifecycle.ts b/src/daemon/lib/acl/lifecycle.ts
new file mode 100644
index 0000000..783dd84
--- /dev/null
+++ b/src/daemon/lib/acl/lifecycle.ts
@@ -0,0 +1,40 @@
+/**
+ * Pure grant-lifecycle logic, extracted from the ACL so it can be unit-tested
+ * without a database and reused verbatim at redeem time and sign time.
+ *
+ * The original #24 bug was possible because redeem-time checked expiry and
+ * sign-time didn't — two definitions of "valid" that drifted. Defining "is
+ * this grant valid right now?" exactly once makes them impossible to disagree.
+ * See aiolabs/nsecbunkerd#25.
+ */
+
+/** The lifecycle fields every grant (Token, SigningCondition) carries. */
+export type Lifecycle = {
+    revokedAt?: Date | null;
+    expiresAt?: Date | null;
+};
+
+/**
+ * "Is this grant valid right now?" — the single lifecycle predicate. A grant
+ * is live iff it has not been revoked and its expiry (if any) is still in the
+ * future. Expiry is treated as exclusive at the boundary: a grant whose
+ * `expiresAt` equals `now` is already dead.
+ */
+export function grantIsLive(grant: Lifecycle, now: Date = new Date()): boolean {
+    if (grant.revokedAt) return false;
+    if (grant.expiresAt && grant.expiresAt.getTime() <= now.getTime()) return false;
+    return true;
+}
+
+/**
+ * `grantIsLive` expressed as a Prisma `where` fragment, so the live filter
+ * runs in the query rather than in app code after the fetch. `now` is threaded
+ * in explicitly so a single request evaluates every row against one clock
+ * reading. Kept in lockstep with `grantIsLive` (see lifecycle.test.ts).
+ */
+export function liveWhere(now: Date) {
+    return {
+        revokedAt: null,
+        OR: [{ expiresAt: null }, { expiresAt: { gt: now } }],
+    };
+}
diff --git a/tests/lifecycle.test.ts b/tests/lifecycle.test.ts
new file mode 100644
index 0000000..4bbe5f1
--- /dev/null
+++ b/tests/lifecycle.test.ts
@@ -0,0 +1,44 @@
+import { test } from 'node:test';
+import assert from 'node:assert/strict';
+import { grantIsLive, liveWhere } from '../src/daemon/lib/acl/lifecycle';
+
+// Fixed reference clock so the assertions don't depend on wall time.
+const now = new Date('2026-06-19T12:00:00.000Z');
+const past = new Date(now.getTime() - 60_000);
+const future = new Date(now.getTime() + 60_000);
+
+test('grantIsLive: no revoke, no expiry -> live', () => {
+    assert.equal(grantIsLive({}, now), true);
+    assert.equal(grantIsLive({ revokedAt: null, expiresAt: null }, now), true);
+});
+
+test('grantIsLive: future expiry -> live', () => {
+    assert.equal(grantIsLive({ expiresAt: future }, now), true);
+});
+
+test('grantIsLive: past expiry -> dead (the #24 case the old code missed at sign time)', () => {
+    assert.equal(grantIsLive({ expiresAt: past }, now), false);
+});
+
+test('grantIsLive: expiry exactly now -> dead (boundary is exclusive)', () => {
+    assert.equal(grantIsLive({ expiresAt: new Date(now.getTime()) }, now), false);
+});
+
+test('grantIsLive: revoked -> dead even with a future expiry (revoke wins)', () => {
+    assert.equal(grantIsLive({ revokedAt: past, expiresAt: future }, now), false);
+});
+
+test('grantIsLive: defaults now to the current time', () => {
+    assert.equal(grantIsLive({ expiresAt: new Date(Date.now() + 3_600_000) }), true);
+    assert.equal(grantIsLive({ expiresAt: new Date(Date.now() - 3_600_000) }), false);
+});
+
+// liveWhere is the SQL mirror of grantIsLive; pin its shape so the two
+// can't silently drift (a drift would re-open the redeem-vs-sign gap #25
+// exists to close).
+test('liveWhere: mirrors grantIsLive as a prisma where-fragment', () => {
+    assert.deepEqual(liveWhere(now), {
+        revokedAt: null,
+        OR: [{ expiresAt: null }, { expiresAt: { gt: now } }],
+    });
+});