diff --git a/BACKEND-API-REFERENCE.md b/BACKEND-API-REFERENCE.md
index e8b735c..28acb31 100644
--- a/BACKEND-API-REFERENCE.md
+++ b/BACKEND-API-REFERENCE.md
@@ -115,6 +115,33 @@ Storage: `localStorage['ed25519AdminToken']` (access), `localStorage['ed25519Adm
 
 **Open decision (business, not technical — ask a human):** whether Mechanism A is retired outright in favor of Mechanism B at cutover, or both run in parallel gated by role/tenant config.
 
+### 2c. Email/phone OTP login — customer (NOT IMPLEMENTED, proposed)
+
+**Gap:** customer storefront login/checkout requires Telegram (Mechanism A) — shoppers without Telegram have no way to identify themselves. Raised as a real usability problem, not a hypothetical.
+
+**Ask:** a third, independent auth mechanism (coexists with 2a/2b, replaces neither):
+
+```
+POST /auth/otp/request
+Body: { "identifier": "user@example.com" }   // or E.164 phone, e.g. "+79991234567"
+Response: { "requestId": "...", "expiresAt": "2026-08-15T10:15:00Z" }
+```
+
+```
+POST /auth/otp/verify
+Body: { "requestId": "...", "code": "482913" }
+Response (on success): {
+  "sessionId": "...", "userId": 8823771, "username": null,
+  "displayName": "user@example.com", "active": true, "expires": "2026-08-15T11:15:00Z"
+}
+```
+
+The success response must be shaped identically to the existing `AuthSession` (`sessionId, userId, username, displayName, active, expires`, §2a's client model) — this lets every existing downstream consumer (guards, session signals, cart/checkout) work unchanged regardless of which mechanism produced the session.
+
+Rate limiting/expiry, explicit so nothing is left to guesswork: 60s resend cooldown per identifier between `/request` calls; code expires 10 minutes after issuance; `requestId` is single-use (a `/verify` call consumes it on success or failure — a fresh `/request` is needed after either).
+
+See `docs/superpowers/specs/2026-08-15-email-phone-login-design.md` for the full design. No client code exists yet — nothing to build against a 404.
+
 ---
 
 ## 3. Bootstrap — the runtime config document
diff --git a/docs/superpowers/specs/2026-08-15-email-phone-login-design.md b/docs/superpowers/specs/2026-08-15-email-phone-login-design.md
new file mode 100644
index 0000000..70d4c20
--- /dev/null
+++ b/docs/superpowers/specs/2026-08-15-email-phone-login-design.md
@@ -0,0 +1,62 @@
+# Email/phone customer login (OTP) — design
+
+**Status:** Approved
+**Date:** 2026-08-15
+**Related backlog item:** #4 (Telegram-only identification)
+
+## Problem
+
+Customer storefront login/checkout requires Telegram today (`src/app/services/auth.service.ts`, `TelegramSessionApiService`) — shoppers without Telegram have no way to identify themselves. User asked for email/phone as an alternative.
+
+Backend has zero email/phone/OTP/password infrastructure — only Telegram session polling exists (`BACKEND-API-REFERENCE.md` §2a). Building real authentication client-side is not possible; this is fundamentally a backend feature. Per user's explicit choice, this round produces the design + backend spec only — no client UI/code, since there is no real backend to build a working feature against yet (mirrors the currency-FX and admin-notifications backend-dependency pattern already documented this session).
+
+## Design
+
+**Mechanism: OTP code (email or SMS)**, chosen over magic link (email-only, extra click) and password (heaviest backend lift — storage, hashing, reset flow). Passwordless matches the feel of the existing Telegram QR flow.
+
+**A third, independent auth mechanism** — coexists with Telegram QR (§2a) and admin Ed25519 (§2b, still unimplemented) exactly the way those two already coexist. Does not replace or modify either.
+
+### Proposed backend endpoints
+
+```
+POST /auth/otp/request
+Body: { "identifier": "user@example.com" }   // or E.164 phone: "+79991234567"
+Response: { "requestId": "...", "expiresAt": "2026-08-15T10:15:00Z" }
+```
+
+```
+POST /auth/otp/verify
+Body: { "requestId": "...", "code": "482913" }
+Response (on success): {
+  "sessionId": "...",
+  "userId": 8823771,
+  "username": null,
+  "displayName": "user@example.com",
+  "active": true,
+  "expires": "2026-08-15T11:15:00Z"
+}
+```
+
+The success response is shaped identically to the existing `AuthSession` model (`src/app/models/auth.model.ts`) — `sessionId`, `userId`, `username`, `displayName`, `active`, `expires`. This is deliberate: every downstream consumer (auth guards, session signals, cart/checkout) already works against `AuthSession` regardless of which mechanism produced it, so wiring this in later requires no changes to guards or session state — only a new "request/verify" UI flow that ends by populating the same session shape Telegram QR already produces.
+
+**Rate limiting / expiry (backend-enforced, not left implicit):**
+- Resend cooldown: 60s between `POST /auth/otp/request` calls for the same identifier.
+- Code expiry: 10 minutes from issuance.
+- `requestId` is single-use — a verify call consumes it regardless of success/failure; a new request is needed after either a wrong code or expiry.
+
+**Identifier validation:** email vs. phone format is auto-detected client-side. Extract the validation logic already written inline in `cart.component.ts` (`validateEmail`/`validatePhone`, currently only used for post-purchase contact capture) into a shared utility rather than duplicating it when the client UI is eventually built — the same email/phone shape-checking applies to both use cases.
+
+### Future client UI (not built this round)
+
+A "Login with email or phone" option next to the existing Telegram QR button: identifier entry → code entry → session established. Deferred until the backend endpoints above exist — no client code to write against a 404.
+
+## Backend doc update
+
+New `BACKEND-API-REFERENCE.md` §2c ("Email/phone OTP login — customer (NOT IMPLEMENTED)"), following the same Gap/Ask format as the existing §12.x entries, documenting the two endpoints, the response-shape compatibility requirement, and the rate-limit/expiry asks above.
+
+## Out of scope
+
+- Admin backoffice login — user confirmed this round is customer-storefront only (item 4 was split into two potential specs during brainstorming; admin auth is a separate future spec if wanted).
+- Magic link and password mechanisms — considered, OTP chosen.
+- Any client-side UI or session-handling code — explicitly deferred; nothing to build against a non-existent backend.
+- Account merging (e.g. a shopper who later links Telegram + email to the same identity) — not raised, not designed.