From 9e084e18a7ed43cc94ca2ba5362ba4dfeaa1e117 Mon Sep 17 00:00:00 2001
From: Diego Rodrigues de Sa e Souza
 <8016841+diegosouzapw@users.noreply.github.com>
Date: Sat, 18 Jul 2026 03:11:01 -0300
Subject: [PATCH] feat: OpenRouter quota tracking (key/credits + free-window
 counter) (#6842) (#7651)

Validated in merge-train --fast @ 4ed4498 (static gates + changed tests 29/29 + vitest green, 2m39s; full suite ran today on trains 1/2c)
---
 .../6842-openrouter-quota-tracking.md         |   1 +
 config/quality/file-size-baseline.json        |   3 +-
 docs/architecture/CODEBASE_DOCUMENTATION.md   |   2 +-
 docs/reference/FREE_TIERS.md                  |   2 +-
 open-sse/config/providerErrorRules.ts         |  22 ++
 open-sse/executors/base.ts                    |  24 ++
 open-sse/services/accountFallback.ts          |  12 +-
 open-sse/services/openrouterFreeWindow.ts     | 255 +++++++++++++
 open-sse/services/openrouterQuotaFetcher.ts   | 357 ++++++++++++++++++
 open-sse/services/usage.ts                    |   4 +
 open-sse/services/usage/openrouter.ts         |  90 +++++
 src/sse/handlers/chat.ts                      |   3 +-
 src/sse/services/auth.ts                      |   7 +-
 ...openrouter-free-window-wiring-6842.test.ts | 223 +++++++++++
 tests/unit/openrouter-quota-6842.test.ts      | 345 +++++++++++++++++
 15 files changed, 1341 insertions(+), 9 deletions(-)
 create mode 100644 changelog.d/features/6842-openrouter-quota-tracking.md
 create mode 100644 open-sse/services/openrouterFreeWindow.ts
 create mode 100644 open-sse/services/openrouterQuotaFetcher.ts
 create mode 100644 open-sse/services/usage/openrouter.ts
 create mode 100644 tests/unit/openrouter-free-window-wiring-6842.test.ts
 create mode 100644 tests/unit/openrouter-quota-6842.test.ts

diff --git a/changelog.d/features/6842-openrouter-quota-tracking.md b/changelog.d/features/6842-openrouter-quota-tracking.md
new file mode 100644
index 0000000000..5cd075a818
--- /dev/null
+++ b/changelog.d/features/6842-openrouter-quota-tracking.md
@@ -0,0 +1 @@
+- **feat(sse):** OpenRouter quota tracking — a dedicated fetcher polls `/api/v1/key` + `/api/v1/credits` (per-key credit cap/remaining/reset, daily/weekly/monthly USD spend, BYOK usage) with a 45s cache and graceful degradation, a local per-account counter tracks the `:free`-model 50-or-1000-per-day + 20 RPM windows (corrected from `X-RateLimit-*` headers and `Retry-After` on 429), and OpenRouter `402` responses now lock the connection with a real cooldown instead of triggering an immediate reselection of the same credit-exhausted key (#6842).
diff --git a/config/quality/file-size-baseline.json b/config/quality/file-size-baseline.json
index d77d9569f9..061dc32e7a 100644
--- a/config/quality/file-size-baseline.json
+++ b/config/quality/file-size-baseline.json
@@ -1,4 +1,5 @@
 {
+  "_rebaseline_2026_07_17_v3849_6842_free_window_wiring": "PR #7651 (openrouter :free-window quota tracking) follow-up: the counter shipped built but never wired into the request pipeline, so combos kept spending guaranteed-429 requests on exhausted free-tier targets. Own growth: src/sse/services/auth.ts 2461->2462 (+1, irreducible at the existing model-aware preflight chokepoint — the `provider === \"codex\"` check that forwards requestedModel into the connection arg is extended to also cover `openrouter`, one added boolean + a doc comment, offset to a single net line by dropping the now-redundant inline condition). Enforcement itself lives in open-sse/services/openrouterQuotaFetcher.ts (not frozen) and the dispatch-time record/correct hooks live in open-sse/executors/base.ts (not frozen). Covered by tests/unit/openrouter-free-window-wiring-6842.test.ts.",
   "_rebaseline_2026_07_17_v3849_ownerprs_media_audio": "Own-growth from the v3.8.49 owner-PR merge campaign (fast-gates PR->release do not run check:file-size, so this surfaced only on re-sync): src/app/api/usage/analytics/route.ts 942->948 (+6 = the 180d/365d getRangeStartIso cases from #7213 usage-extended-periods) and tests/unit/audio-transcription-handler.test.ts new testFrozen 824 (Gladia async STT test cases added by #7603). Both irreducible additions covered by their PR tests; structural shrink tracked in #3501.",
   "_rebaseline_2026_07_14_7034_goog_api_key": "Issue #7034 (gemini-cli x-goog-api-key client auth) own growth: src/sse/services/auth.ts 2458->2461 (+3 = import + the two-line extractGoogApiKeyHeader() call/return at the existing extractApiKey() chokepoint, plus a 1-line doc-comment mention offset by a 1-line net save elsewhere in the same edit). The actual header-read/trim logic was EXTRACTED into a new leaf module src/sse/services/googApiKeyAuth.ts (shared by both extractApiKey() here and extractBearer() in src/server/authz/policies/clientApi.ts, which is not frozen) to keep this frozen file's growth to the irreducible call-site wiring. Covered by tests/unit/auth-extract-api-key.test.ts and tests/unit/authz/client-api-policy.test.ts.",
   "_rebaseline_2026_07_14_6928_comfyui_baseurl_override": "Issue #6928 own growth: open-sse/handlers/videoGeneration.ts 1265->1275 (+10 = resolveComfyUiBaseUrl import + expanding the comfyui dispatch call into a multi-line object literal so the per-connection providerSpecificData.baseUrl override — same storage convention self-hosted chat providers use — is threaded through to handleComfyUIVideoGeneration; Prettier's 100-char width forces the multi-line form), src/app/(dashboard)/dashboard/providers/[id]/providerPageHelpers.ts 1053->1054 (+1 = comfyui added to CONFIGURABLE_BASE_URL_PROVIDERS/DEFAULT_PROVIDER_BASE_URLS/getProviderBaseUrlPlaceholder so the Add/Edit connection modals render an editable base-URL field for ComfyUI, mirroring self-hosted chat providers). The identical dispatch pattern was also applied to imageGeneration.ts and musicGeneration.ts, both well under their frozen caps. Covered by tests/unit/comfyui-baseurl-override-6928.test.ts (resolver unit tests + handler-level fetch-mock overrides for image/video/music) and the new provider-page-helpers-3501.test.ts assertion.",
@@ -270,7 +271,7 @@
     "_rebaseline_2026_07_09_6678_routing_strategy_9router": "#6678 (SeaXen) — 9router-parity Routing Strategy settings card + per-provider/combo sticky-round-robin override. Own growth: ProviderDetailPageClient.tsx 784->786 (single ProviderAccountRoutingCard mount + import), auth.ts 2448->2458 (providerStrategies override resolution: fallbackStrategy/stickyRoundRobinLimit per-provider cascade in getProviderCredentials). Both additive, zero unrelated refactor; new UI/logic lives in new files (ProviderAccountRoutingCard.tsx, RoutingStrategyCard.tsx, rrState.ts::resolveComboStickyRoundRobinLimit). chat.ts value below reflects the current release tip (grown by other concurrent PRs, e.g. #6640), not this PR own change.",
     "src/sse/handlers/chat.ts": 1796,
     "src/sse/handlers/chatHelpers.ts": 876,
-    "src/sse/services/auth.ts": 2461,
+    "src/sse/services/auth.ts": 2462,
     "open-sse/executors/default.ts": 877,
     "open-sse/translator/request/openai-responses.ts": 902,
     "open-sse/executors/kiro.ts": 944,
diff --git a/docs/architecture/CODEBASE_DOCUMENTATION.md b/docs/architecture/CODEBASE_DOCUMENTATION.md
index ccf2ea0d36..513b144ee0 100644
--- a/docs/architecture/CODEBASE_DOCUMENTATION.md
+++ b/docs/architecture/CODEBASE_DOCUMENTATION.md
@@ -527,7 +527,7 @@ Highlights (full list under `open-sse/services/`):
 | Combo routing        | `combo.ts` (17 strategies), `comboConfig.ts`, `comboMetrics.ts`, `comboManifestMetrics.ts`, `comboAgentMiddleware.ts`                                                                                                                             |
 | Auto Combo engine    | `autoCombo/` — `engine.ts`, `scoring.ts`, `taskFitness.ts`, `virtualFactory.ts`, `modePacks.ts`, `autoPrefix.ts`, `persistence.ts`, `providerDiversity.ts`, `providerRegistryAccessor.ts`, `routerStrategy.ts`, `selfHealing.ts`, `index.ts`      |
 | Resilience           | `accountFallback.ts` (cooldown + lockout), `errorClassifier.ts`, `emergencyFallback.ts`, `rateLimitManager.ts`, `rateLimitSemaphore.ts`, `accountSemaphore.ts`, `accountSelector.ts`                                                              |
-| Quotas               | `quotaMonitor.ts`, `quotaPreflight.ts`, `bailianQuotaFetcher.ts`, `codexQuotaFetcher.ts`, `deepseekQuotaFetcher.ts`, `crofUsageFetcher.ts`, `antigravityCredits.ts`                                                                               |
+| Quotas               | `quotaMonitor.ts`, `quotaPreflight.ts`, `bailianQuotaFetcher.ts`, `codexQuotaFetcher.ts`, `deepseekQuotaFetcher.ts`, `openrouterQuotaFetcher.ts`, `openrouterFreeWindow.ts`, `crofUsageFetcher.ts`, `antigravityCredits.ts`                       |
 | Caching              | `reasoningCache.ts`, `searchCache.ts`, `signatureCache.ts`, `requestDedup.ts`                                                                                                                                                                     |
 | Routing intelligence | `intentClassifier.ts`, `taskAwareRouter.ts`, `backgroundTaskDetector.ts`, `volumeDetector.ts`, `wildcardRouter.ts`, `workflowFSM.ts`, `specificityDetector.ts`, `specificityRules.ts`, `specificityTypes.ts`                                      |
 | Model handling       | `modelCapabilities.ts`, `modelDeprecation.ts`, `modelFamilyFallback.ts`, `modelStrip.ts`, `model.ts`, `provider.ts`, `providerRequestDefaults.ts`, `providerCostData.ts`, `payloadRules.ts`                                                       |
diff --git a/docs/reference/FREE_TIERS.md b/docs/reference/FREE_TIERS.md
index 775586ab0a..936af9e78c 100644
--- a/docs/reference/FREE_TIERS.md
+++ b/docs/reference/FREE_TIERS.md
@@ -299,7 +299,7 @@ A 50-agent web-research pass (official docs + last-7-days news, adversarially ve
 - **`nous-research`** — The shipped freeNote ("Free tier: 50 RPM, 500,000 TPM") does not match the current Nous Portal product. The portal launched April 27, 2026 and structures its free tier as $0.10/month in recurring cre…
 - **`nvidia`** — The "40 RPM, 70+ models" rate limit element matches the catalog, but the freeNote framing as a simple dev-access tier undersells that the old one-time credit pool has been removed — access is now tru…
 - **`ollama-cloud`** — Our shipped freeNote is "(none)" — this is stale. Ollama Cloud launched a cloud inference product with a genuine free tier that provides light weekly GPU-time-based access to hosted open models.
-- **`openrouter`** — RPD tightened from 200 to 50 for zero-credit accounts (RPM unchanged at 20). The catalog note was accurate on RPM but overstated the RPD by 4x for the no-credits baseline tier.
+- **`openrouter`** — RPD tightened from 200 to 50 for zero-credit accounts (RPM unchanged at 20). The catalog note was accurate on RPM but overstated the RPD by 4x for the no-credits baseline tier. **Runtime tracking (#6842)**: this is no longer just a static note — `open-sse/services/openrouterQuotaFetcher.ts` polls `/api/v1/key` + `/api/v1/credits` for per-key credit cap/remaining/reset and daily/weekly/monthly USD spend, and `open-sse/services/openrouterFreeWindow.ts` locally tracks the `:free`-model 50-or-1000-per-day + 20 RPM windows described above (corrected from `X-RateLimit-*` response headers on 429s), surfaced in Dashboard → Provider Quota.
 - **`phind`** — Phind shut down on January 16, 2026. The provider has now been **fully removed** from the catalog (registry, executor, and both the web-cookie and API-key catalog entries) — matching the dead-service-removal precedent (#5246 Gemini CLI).
 - **`pollinations`** — Partially matches — the "no API key required" claim is still true for anonymous access, but the catalog freeNote omits that: (1) rate limits do apply (interval throttle of ~1 req/6-15s for anonymous …
 - **`predibase`** — The shipped freeNote ($25 free trial credits, 30-day validity) still matches current documentation. However, the catalog omits the concurrent 20,000 tokens/day serverless rate limit that applies duri…
diff --git a/open-sse/config/providerErrorRules.ts b/open-sse/config/providerErrorRules.ts
index dde56a53d2..ce5c74e702 100644
--- a/open-sse/config/providerErrorRules.ts
+++ b/open-sse/config/providerErrorRules.ts
@@ -155,6 +155,27 @@ function buildCloudflareAiRules(): ProviderErrorRule[] {
   ];
 }
 
+// ─── OpenRouter ─────────────────────────────────────────────────────────────
+// #6842: OpenRouter returns 402 for both a negative account balance and a
+// depleted per-key credit cap. The global `status_402` rule already maps this
+// to `quota_exhausted` with a zero cooldown (immediate fallback to the next
+// connection), but leaves the scope ambiguous and doesn't stop the SAME
+// connection from being reselected instantly (credits genuinely need a
+// top-up, not a timed wait). This explicit rule locks the whole connection
+// (scope: "connection" — credits are account-wide, not per-model) for a real
+// cooldown so combo routing skips it instead of hot-looping back onto it.
+function buildOpenrouterRules(): ProviderErrorRule[] {
+  return [
+    {
+      id: "openrouter-credit-exhausted-402",
+      match: ({ status }) => {
+        if (status !== 402) return null;
+        return { reason: "quota_exhausted", scope: "connection", cooldownMs: 2 * 60 * 1000 };
+      },
+    },
+  ];
+}
+
 /**
  * Global registry. Provider name → ordered list of rules (first match wins).
  * Add new providers here; the matcher in classifyError will pick them up
@@ -167,6 +188,7 @@ export const providerRuleRegistry = new Map<string, ProviderErrorRule[]>([
   ["minimax", buildMinimaxRules()],
   ["minimax-passthrough", buildMinimaxRules()],
   ["cloudflare-ai", buildCloudflareAiRules()],
+  ["openrouter", buildOpenrouterRules()],
 ]);
 
 /**
diff --git a/open-sse/executors/base.ts b/open-sse/executors/base.ts
index 3d82de7680..6b4b5d841c 100644
--- a/open-sse/executors/base.ts
+++ b/open-sse/executors/base.ts
@@ -17,6 +17,12 @@ import {
 import { applyFingerprint, isCliCompatEnabled } from "../config/cliFingerprints.ts";
 import { supportsClaudeMaxEffort, supportsXHighEffort } from "../config/providerModels.ts";
 import { getThinkingBudgetConfig, ThinkingMode } from "../services/thinkingBudget.ts";
+import {
+  recordFreeWindowAttempt,
+  correctFromRateLimitHeaders,
+  resolveAccountKey,
+  isFreeVariantModel,
+} from "../services/openrouterFreeWindow.ts";
 import type { PoolConfig } from "../services/sessionPool/types.ts";
 import type { Session } from "../services/sessionPool/session.ts";
 import { SessionPool } from "../services/sessionPool/sessionPool.ts";
@@ -1214,8 +1220,26 @@ export class BaseExecutor {
           body: bodyString,
         };
 
+        // OpenRouter `:free`-variant local window (#6842): record every real
+        // dispatch attempt (failed attempts still consume a request slot per
+        // OpenRouter's own accounting) and self-correct the local counters
+        // from the upstream `X-RateLimit-*` headers on the response. Scoped
+        // to `:free` models only — no-op (and no extra work) for every other
+        // OpenRouter request or provider.
+        const openrouterFreeWindowAccountKey =
+          this.provider === "openrouter" && isFreeVariantModel(model) && activeCredentials.connectionId
+            ? resolveAccountKey(activeCredentials.connectionId, activeCredentials)
+            : null;
+        if (openrouterFreeWindowAccountKey) {
+          recordFreeWindowAttempt(openrouterFreeWindowAccountKey);
+        }
+
         let response = await fetchWithStartTimeout(url, fetchOptions);
 
+        if (openrouterFreeWindowAccountKey) {
+          correctFromRateLimitHeaders(openrouterFreeWindowAccountKey, response.headers);
+        }
+
         // Context Editing 400-fallback for Claude-compatible relays.
         if (
           response.status === HTTP_STATUS.BAD_REQUEST &&
diff --git a/open-sse/services/accountFallback.ts b/open-sse/services/accountFallback.ts
index 32562fc526..10347ec0f3 100644
--- a/open-sse/services/accountFallback.ts
+++ b/open-sse/services/accountFallback.ts
@@ -1554,13 +1554,21 @@ export function checkFallbackError(
       }
       return fallback;
     }
-    const cooldownMs = configuredRule.cooldownMs ?? 0;
+    // #6842: non-backoff configured rules (e.g. status_402) previously never
+    // consulted providerRuleRegistry, so a provider-specific rule (like
+    // OpenRouter's credit-exhausted 402 lock) could never override the
+    // generic zero-cooldown default. Mirror the backoff branch above so
+    // provider rules win on cooldown/reason regardless of `backoff`.
+    const providerMatch = provider
+      ? getProviderErrorRuleMatch(provider, status, headers, structuredError ?? null)
+      : null;
+    const cooldownMs = providerMatch?.cooldownMs ?? configuredRule.cooldownMs ?? 0;
     return {
       shouldFallback: true,
       cooldownMs,
       baseCooldownMs: cooldownMs,
       configuredCooldownMs: cooldownMs,
-      reason: configuredRule.reason ?? RateLimitReason.UNKNOWN,
+      reason: providerMatch?.reason ?? configuredRule.reason ?? RateLimitReason.UNKNOWN,
     };
   }
 
diff --git a/open-sse/services/openrouterFreeWindow.ts b/open-sse/services/openrouterFreeWindow.ts
new file mode 100644
index 0000000000..4f05061a18
--- /dev/null
+++ b/open-sse/services/openrouterFreeWindow.ts
@@ -0,0 +1,255 @@
+/**
+ * openrouterFreeWindow.ts — OpenRouter `:free`-variant local window tracker (#6842)
+ *
+ * OpenRouter's official monitoring API (`/api/v1/key`, `/api/v1/credits`) reports
+ * USD spend, never request counts — so the `:free`-model per-account request
+ * windows (docs/reference/FREE_TIERS.md) cannot be read from those endpoints.
+ * This module tracks them locally instead:
+ *
+ *   - A UTC-day counter: 50 requests/day at $0 all-time purchased, 1000/day
+ *     once $10+ has been purchased (operator-overridable via setPurchasedTier).
+ *   - A 20 RPM rolling window (true rolling — timestamps pruned to the last 60s,
+ *     not a fixed-bucket reset).
+ *
+ * Bucketed by ACCOUNT, not by connection/key — multiple OmniRoute connections
+ * that share one upstream OpenRouter account must share one window. OmniRoute's
+ * `provider_connections` are per-key, so callers resolve an account bucket key
+ * via `resolveAccountKey()`: an explicit `providerSpecificData.openrouterAccountKey`
+ * groups keys under one account; otherwise each connection gets its own bucket
+ * (safe default — never worse than no tracking).
+ *
+ * State is in-memory only (module-level Map, not persisted). This is a
+ * deliberate MVP scope: local counting is inherently best-effort (drifts on
+ * process restart or multi-instance deployments sharing one OpenRouter
+ * account) and is corrected from upstream `X-RateLimit-*` response headers on
+ * every 429, which is authoritative. See the plan's "Risks" section — SQLite
+ * persistence is a possible follow-up, not required for correctness here.
+ */
+
+const RPM_WINDOW_MS = 60_000;
+const RPM_LIMIT = 20;
+const DAILY_LIMIT_BASE = 50;
+const DAILY_LIMIT_PURCHASED = 1000;
+
+interface AccountWindowState {
+  dayKey: string;
+  dayCount: number;
+  purchasedAtLeast10: boolean;
+  requestTimestamps: number[];
+  serverDailyLimit: number | null;
+  serverDailyRemaining: number | null;
+  serverResetAtMs: number | null;
+}
+
+export interface FreeWindowStatus {
+  dailyLimit: number;
+  dailyUsed: number;
+  dailyRemaining: number;
+  dailyResetAt: string;
+  rpmLimit: number;
+  rpmUsed: number;
+  rpmRemaining: number;
+}
+
+const accountWindows = new Map<string, AccountWindowState>();
+
+function utcDayKey(now: number): string {
+  return new Date(now).toISOString().slice(0, 10);
+}
+
+function nextUtcMidnightIso(now: number): string {
+  const date = new Date(now);
+  const next = Date.UTC(
+    date.getUTCFullYear(),
+    date.getUTCMonth(),
+    date.getUTCDate() + 1,
+    0,
+    0,
+    0,
+    0
+  );
+  return new Date(next).toISOString();
+}
+
+/**
+ * Whether a model id is an OpenRouter `:free`-variant (e.g.
+ * `x-ai/grok-4-fast:free`). Shared by the dispatch-time record/correct hooks
+ * (`open-sse/executors/base.ts`) and the quota-preflight enforcement hook
+ * (`open-sse/services/openrouterQuotaFetcher.ts`) so both sides agree on the
+ * same definition of "free variant."
+ */
+export function isFreeVariantModel(model: string | null | undefined): boolean {
+  return typeof model === "string" && model.endsWith(":free");
+}
+
+/**
+ * Resolve the shared account bucket key for a connection. Operators group
+ * multiple keys under one OpenRouter account via `providerSpecificData.
+ * openrouterAccountKey`; without it, each connection is its own bucket.
+ */
+export function resolveAccountKey(
+  connectionId: string,
+  connection?: Record<string, unknown> | null
+): string {
+  const psd = connection?.providerSpecificData as Record<string, unknown> | undefined;
+  const explicit = typeof psd?.openrouterAccountKey === "string" ? psd.openrouterAccountKey : "";
+  return explicit.trim().length > 0 ? `acct:${explicit.trim()}` : `conn:${connectionId}`;
+}
+
+function getOrInitState(accountKey: string, now: number): AccountWindowState {
+  const dayKey = utcDayKey(now);
+  const existing = accountWindows.get(accountKey);
+  if (existing && existing.dayKey === dayKey) return existing;
+
+  const fresh: AccountWindowState = {
+    dayKey,
+    dayCount: 0,
+    purchasedAtLeast10: existing?.purchasedAtLeast10 ?? false,
+    requestTimestamps: [],
+    serverDailyLimit: null,
+    serverDailyRemaining: null,
+    serverResetAtMs: null,
+  };
+  accountWindows.set(accountKey, fresh);
+  return fresh;
+}
+
+function pruneRpmWindow(state: AccountWindowState, now: number): void {
+  const cutoff = now - RPM_WINDOW_MS;
+  state.requestTimestamps = state.requestTimestamps.filter((ts) => ts > cutoff);
+}
+
+/**
+ * Operator override: declare whether $10+ has been purchased all-time on this
+ * account, unlocking the 1000/day tier instead of the 50/day default.
+ */
+export function setPurchasedTier(accountKey: string, purchasedAtLeast10: boolean): void {
+  const state = getOrInitState(accountKey, Date.now());
+  state.purchasedAtLeast10 = purchasedAtLeast10;
+}
+
+/**
+ * Record a `:free`-variant request attempt against the account bucket.
+ * Failed attempts count toward the daily cap too (per OpenRouter's own
+ * accounting — a rejected request still consumed a request slot).
+ */
+export function recordFreeWindowAttempt(accountKey: string, now: number = Date.now()): void {
+  const state = getOrInitState(accountKey, now);
+  pruneRpmWindow(state, now);
+  state.dayCount += 1;
+  state.requestTimestamps.push(now);
+}
+
+function getHeader(headers: Headers | Record<string, string>, name: string): string | null {
+  if (typeof (headers as Headers).get === "function") {
+    return (headers as Headers).get(name);
+  }
+  const record = headers as Record<string, string>;
+  return record[name] ?? record[name.toLowerCase()] ?? null;
+}
+
+function parseResetMsFromHeader(reset: string | null): number | null {
+  if (reset === null) return null;
+  const resetNum = Number(reset);
+  if (!Number.isFinite(resetNum)) return null;
+  return resetNum > 10_000_000_000 ? resetNum : resetNum * 1000;
+}
+
+function parseRetryAfterMs(retryAfter: string | null, now: number): number | null {
+  if (retryAfter === null) return null;
+  const seconds = Number(retryAfter);
+  if (!Number.isFinite(seconds) || seconds <= 0) return null;
+  return now + seconds * 1000;
+}
+
+function resolveResetMs(
+  reset: string | null,
+  retryAfter: string | null,
+  now: number
+): number | null {
+  const headerResetMs = parseResetMsFromHeader(reset);
+  const retryAfterMs = parseRetryAfterMs(retryAfter, now);
+  if (retryAfterMs === null) return headerResetMs;
+  return headerResetMs === null ? retryAfterMs : Math.max(headerResetMs, retryAfterMs);
+}
+
+/**
+ * Correct local counters from OpenRouter's authoritative rate-limit headers,
+ * present on 429 responses per OpenRouter's docs. `Retry-After` (seconds)
+ * is folded into the reset timestamp when present and later than the
+ * `X-RateLimit-Reset` value.
+ */
+export function correctFromRateLimitHeaders(
+  accountKey: string,
+  headers: Headers | Record<string, string>,
+  now: number = Date.now()
+): void {
+  const state = getOrInitState(accountKey, now);
+  const limit = getHeader(headers, "x-ratelimit-limit");
+  const remaining = getHeader(headers, "x-ratelimit-remaining");
+
+  if (limit !== null && Number.isFinite(Number(limit))) {
+    state.serverDailyLimit = Number(limit);
+  }
+  if (remaining !== null && Number.isFinite(Number(remaining))) {
+    state.serverDailyRemaining = Number(remaining);
+  }
+
+  const resetMs = resolveResetMs(
+    getHeader(headers, "x-ratelimit-reset"),
+    getHeader(headers, "retry-after"),
+    now
+  );
+  if (resetMs !== null) {
+    state.serverResetAtMs = resetMs;
+  }
+}
+
+function resolveDailyLimit(state: AccountWindowState): number {
+  if (state.serverDailyLimit !== null) return state.serverDailyLimit;
+  return state.purchasedAtLeast10 ? DAILY_LIMIT_PURCHASED : DAILY_LIMIT_BASE;
+}
+
+function resolveDailyUsed(state: AccountWindowState, dailyLimit: number): number {
+  if (state.serverDailyRemaining !== null) {
+    return Math.max(0, dailyLimit - state.serverDailyRemaining);
+  }
+  return state.dayCount;
+}
+
+/**
+ * Current window status for the account bucket: daily count vs limit
+ * (50-or-1000, server-corrected when available) and the 20 RPM rolling
+ * window, plus reset timestamps for the dashboard countdown.
+ */
+export function getFreeWindowStatus(
+  accountKey: string,
+  now: number = Date.now()
+): FreeWindowStatus {
+  const state = getOrInitState(accountKey, now);
+  pruneRpmWindow(state, now);
+
+  const dailyLimit = resolveDailyLimit(state);
+  const dailyUsed = resolveDailyUsed(state, dailyLimit);
+  const dailyResetAt =
+    state.serverResetAtMs !== null
+      ? new Date(state.serverResetAtMs).toISOString()
+      : nextUtcMidnightIso(now);
+
+  const rpmUsed = state.requestTimestamps.length;
+
+  return {
+    dailyLimit,
+    dailyUsed,
+    dailyRemaining: Math.max(0, dailyLimit - dailyUsed),
+    dailyResetAt,
+    rpmLimit: RPM_LIMIT,
+    rpmUsed,
+    rpmRemaining: Math.max(0, RPM_LIMIT - rpmUsed),
+  };
+}
+
+/** Test/ops helper — clears all in-memory account window state. */
+export function clearFreeWindowState(): void {
+  accountWindows.clear();
+}
diff --git a/open-sse/services/openrouterQuotaFetcher.ts b/open-sse/services/openrouterQuotaFetcher.ts
new file mode 100644
index 0000000000..136e0d16aa
--- /dev/null
+++ b/open-sse/services/openrouterQuotaFetcher.ts
@@ -0,0 +1,357 @@
+/**
+ * openrouterQuotaFetcher.ts — OpenRouter Key/Credits Quota Fetcher (#6842)
+ *
+ * Implements QuotaFetcher for the OpenRouter provider (quotaPreflight.ts + quotaMonitor.ts).
+ *
+ * OpenRouter exposes two official, documented monitoring endpoints
+ * (https://openrouter.ai/docs/api/reference/limits):
+ *
+ *   GET https://openrouter.ai/api/v1/key
+ *     -> { data: { limit, limit_remaining, limit_reset, usage, usage_daily,
+ *                  usage_weekly, usage_monthly, is_free_tier, byok_usage,
+ *                  include_byok_in_limit } }
+ *     `limit`/`limit_remaining` are per-key USD credit caps — null means
+ *     unlimited/never set. `limit_reset` is null when the cap never resets.
+ *
+ *   GET https://openrouter.ai/api/v1/credits
+ *     -> { data: { total_credits, total_usage } }
+ *     Account-level totals; upstream caches this endpoint for ~60s already.
+ *
+ * We fetch both (credits is a cheap second call, same auth) and merge into one
+ * QuotaInfo. Graceful "unknown" on any fetch failure — quota tracking must
+ * never block routing (mirrors deepseekQuotaFetcher.ts / bailianQuotaFetcher.ts).
+ *
+ * Cache: in-memory TTL (45s, inside the 30-60s window OpenRouter's own docs
+ * recommend) keyed by connectionId, so combo preflight/monitor polling doesn't
+ * hammer the upstream on every request.
+ *
+ * Registration: call registerOpenrouterQuotaFetcher() once at server startup.
+ */
+
+import { registerQuotaFetcher, type QuotaInfo } from "./quotaPreflight.ts";
+import { registerMonitorFetcher } from "./quotaMonitor.ts";
+import { throttleQuotaFetch } from "./quotaFetchThrottle.ts";
+import {
+  getFreeWindowStatus,
+  isFreeVariantModel,
+  resolveAccountKey,
+  type FreeWindowStatus,
+} from "./openrouterFreeWindow.ts";
+
+const OPENROUTER_CONFIG = {
+  baseUrl: "https://openrouter.ai/api/v1",
+  keyPath: "/key",
+  creditsPath: "/credits",
+};
+
+// Cache TTL — inside OpenRouter's documented 30-60s window.
+const CACHE_TTL_MS = 45_000;
+
+export interface OpenrouterQuota extends QuotaInfo {
+  limit: number | null;
+  limitRemaining: number | null;
+  isFreeTier: boolean;
+  usage: number;
+  usageDaily: number;
+  usageWeekly: number;
+  usageMonthly: number;
+  byokUsage: number | null;
+  includeByokInLimit: boolean;
+  totalCredits: number | null;
+  totalUsage: number | null;
+  creditBalance: number | null;
+}
+
+interface CacheEntry {
+  quota: OpenrouterQuota;
+  fetchedAt: number;
+}
+
+const quotaCache = new Map<string, CacheEntry>();
+
+const _cacheCleanup = setInterval(() => {
+  const now = Date.now();
+  for (const [key, entry] of quotaCache) {
+    if (now - entry.fetchedAt > CACHE_TTL_MS * 5) {
+      quotaCache.delete(key);
+    }
+  }
+}, 5 * 60_000);
+
+if (typeof _cacheCleanup === "object" && "unref" in _cacheCleanup) {
+  (_cacheCleanup as { unref?: () => void }).unref?.();
+}
+
+// ─── Helpers ─────────────────────────────────────────────────────────────────
+
+function toRecord(value: unknown): Record<string, unknown> {
+  return value && typeof value === "object" && !Array.isArray(value)
+    ? (value as Record<string, unknown>)
+    : {};
+}
+
+function toNullableNumber(value: unknown): number | null {
+  if (typeof value === "number" && Number.isFinite(value)) return value;
+  if (typeof value === "string" && value.trim().length > 0) {
+    const parsed = Number(value);
+    if (Number.isFinite(parsed)) return parsed;
+  }
+  return null;
+}
+
+function toFiniteNumber(value: unknown, fallback = 0): number {
+  const n = toNullableNumber(value);
+  return n === null ? fallback : n;
+}
+
+function toIsoOrNull(value: unknown): string | null {
+  const n = toNullableNumber(value);
+  if (n === null) return null;
+  const date = new Date(n < 1e12 ? n * 1000 : n);
+  if (Number.isNaN(date.getTime()) || date.getTime() <= 0) return null;
+  return date.toISOString();
+}
+
+// ─── Response Parsers ────────────────────────────────────────────────────────
+
+interface OpenrouterKeyFields {
+  limit: number | null;
+  limitRemaining: number | null;
+  limitReset: string | null;
+  isFreeTier: boolean;
+  usage: number;
+  usageDaily: number;
+  usageWeekly: number;
+  usageMonthly: number;
+  byokUsage: number | null;
+  includeByokInLimit: boolean;
+}
+
+/**
+ * Parse the `GET /api/v1/key` response body. Returns null when the payload
+ * doesn't carry a recognizable `data` object (e.g. an unexpected shape).
+ */
+export function parseOpenrouterKeyResponse(data: unknown): OpenrouterKeyFields | null {
+  const outer = toRecord(data);
+  const inner = "data" in outer ? toRecord(outer.data) : outer;
+  if (Object.keys(inner).length === 0) return null;
+
+  return {
+    limit: toNullableNumber(inner.limit),
+    limitRemaining: toNullableNumber(inner.limit_remaining),
+    limitReset: toIsoOrNull(inner.limit_reset),
+    isFreeTier: inner.is_free_tier === true,
+    usage: toFiniteNumber(inner.usage, 0),
+    usageDaily: toFiniteNumber(inner.usage_daily, 0),
+    usageWeekly: toFiniteNumber(inner.usage_weekly, 0),
+    usageMonthly: toFiniteNumber(inner.usage_monthly, 0),
+    byokUsage: toNullableNumber(inner.byok_usage),
+    includeByokInLimit: inner.include_byok_in_limit === true,
+  };
+}
+
+interface OpenrouterCreditsFields {
+  totalCredits: number | null;
+  totalUsage: number | null;
+}
+
+/**
+ * Parse the `GET /api/v1/credits` response body. Returns nulls (not a full
+ * null-object) when the payload is missing — credits is a best-effort
+ * secondary signal, the key endpoint alone is enough to build a quota.
+ */
+export function parseOpenrouterCreditsResponse(data: unknown): OpenrouterCreditsFields {
+  const outer = toRecord(data);
+  const inner = "data" in outer ? toRecord(outer.data) : outer;
+  return {
+    totalCredits: toNullableNumber(inner.total_credits),
+    totalUsage: toNullableNumber(inner.total_usage),
+  };
+}
+
+function buildQuotaFromParts(
+  key: OpenrouterKeyFields,
+  credits: OpenrouterCreditsFields
+): OpenrouterQuota {
+  const hasCap = key.limit !== null && key.limitRemaining !== null;
+  const limitReached = hasCap && (key.limitRemaining as number) <= 0;
+  const percentUsed = hasCap && key.limit! > 0 ? 1 - key.limitRemaining! / key.limit! : 0;
+  const creditBalance =
+    key.limitRemaining !== null
+      ? key.limitRemaining
+      : credits.totalCredits !== null && credits.totalUsage !== null
+        ? credits.totalCredits - credits.totalUsage
+        : null;
+
+  return {
+    used: percentUsed * 100,
+    total: 100,
+    percentUsed,
+    resetAt: key.limitReset,
+    limitReached,
+    limit: key.limit,
+    limitRemaining: key.limitRemaining,
+    isFreeTier: key.isFreeTier,
+    usage: key.usage,
+    usageDaily: key.usageDaily,
+    usageWeekly: key.usageWeekly,
+    usageMonthly: key.usageMonthly,
+    byokUsage: key.byokUsage,
+    includeByokInLimit: key.includeByokInLimit,
+    totalCredits: credits.totalCredits,
+    totalUsage: credits.totalUsage,
+    creditBalance,
+  };
+}
+
+// ─── Free-Window Preflight (#6842) ───────────────────────────────────────────
+
+/**
+ * Build a `limitReached` QuotaInfo from the local `:free`-window daily
+ * counter — no upstream I/O, so this is safe to call on every preflight
+ * without adding latency or spending a request.
+ */
+function buildFreeWindowExhaustedQuota(status: FreeWindowStatus): QuotaInfo {
+  const percentUsed = status.dailyLimit > 0 ? status.dailyUsed / status.dailyLimit : 1;
+  return {
+    used: status.dailyUsed,
+    total: status.dailyLimit,
+    percentUsed: Math.min(1, Math.max(0, percentUsed)),
+    resetAt: status.dailyResetAt,
+    limitReached: true,
+  };
+}
+
+/**
+ * When the requested model is a `:free` variant and the locally-tracked
+ * daily window is already exhausted, short-circuit before any network call:
+ * the upstream `/key` + `/credits` fetch below only reports USD spend, never
+ * the `:free` request count, so it cannot see this exhaustion on its own —
+ * and dispatching the chat request itself would just spend a guaranteed 429.
+ */
+function checkFreeWindowExhausted(
+  connectionId: string,
+  connection: Record<string, unknown> | undefined,
+  requestedModel: unknown
+): QuotaInfo | null {
+  if (!isFreeVariantModel(typeof requestedModel === "string" ? requestedModel : null)) {
+    return null;
+  }
+  const accountKey = resolveAccountKey(connectionId, connection);
+  const status = getFreeWindowStatus(accountKey);
+  return status.dailyRemaining <= 0 ? buildFreeWindowExhaustedQuota(status) : null;
+}
+
+// ─── Core Fetcher ────────────────────────────────────────────────────────────
+
+async function fetchJson(
+  url: string,
+  apiKey: string
+): Promise<{ status: number; data: unknown } | null> {
+  try {
+    const response = await fetch(url, {
+      method: "GET",
+      headers: {
+        Authorization: `Bearer ${apiKey}`,
+        Accept: "application/json",
+      },
+      signal: AbortSignal.timeout(8_000),
+    });
+    if (!response.ok) return { status: response.status, data: null };
+    const data = await response.json();
+    return { status: response.status, data };
+  } catch {
+    return null;
+  }
+}
+
+/**
+ * Fetch current quota for an OpenRouter connection.
+ * Returns quota info based on the /key + /credits API responses.
+ *
+ * @param connectionId - Connection ID from the DB (used for cache keying)
+ * @param connection - Optional connection object with apiKey
+ * @returns OpenrouterQuota or null if fetch fails / no credentials
+ */
+export async function fetchOpenrouterQuota(
+  connectionId: string,
+  connection?: Record<string, unknown>
+): Promise<QuotaInfo | null> {
+  const cached = quotaCache.get(connectionId);
+  if (cached && Date.now() - cached.fetchedAt < CACHE_TTL_MS) {
+    return cached.quota;
+  }
+
+  const apiKey =
+    typeof connection?.apiKey === "string" && connection.apiKey.trim().length > 0
+      ? connection.apiKey
+      : null;
+  if (!apiKey) return null;
+
+  try {
+    await throttleQuotaFetch();
+
+    const keyUrl = `${OPENROUTER_CONFIG.baseUrl}${OPENROUTER_CONFIG.keyPath}`;
+    const keyResult = await fetchJson(keyUrl, apiKey);
+
+    // 401/403 on the key endpoint: token invalid — remove from cache, fail open.
+    if (!keyResult || keyResult.status === 401 || keyResult.status === 403) {
+      quotaCache.delete(connectionId);
+      return null;
+    }
+    if (keyResult.status !== 200) return null;
+
+    const keyFields = parseOpenrouterKeyResponse(keyResult.data);
+    if (!keyFields) return null;
+
+    const creditsUrl = `${OPENROUTER_CONFIG.baseUrl}${OPENROUTER_CONFIG.creditsPath}`;
+    const creditsResult = await fetchJson(creditsUrl, apiKey);
+    const creditsFields =
+      creditsResult && creditsResult.status === 200
+        ? parseOpenrouterCreditsResponse(creditsResult.data)
+        : { totalCredits: null, totalUsage: null };
+
+    const quota = buildQuotaFromParts(keyFields, creditsFields);
+    quotaCache.set(connectionId, { quota, fetchedAt: Date.now() });
+    return quota;
+  } catch {
+    // Network error, timeout, etc. — fail open (graceful "unknown").
+    return null;
+  }
+}
+
+// ─── Invalidation ────────────────────────────────────────────────────────────
+
+export function invalidateOpenrouterQuotaCache(connectionId: string): void {
+  quotaCache.delete(connectionId);
+}
+
+/**
+ * The fetcher actually wired into quotaPreflight.ts / quotaMonitor.ts (#6842
+ * follow-up). Kept as a thin wrapper — rather than inlined into
+ * fetchOpenrouterQuota() above — so the /key + /credits fetcher itself stays
+ * a plain, independently-testable function and the free-window short-circuit
+ * doesn't add branching to its already-tight complexity budget.
+ */
+export async function fetchOpenrouterQuotaWithFreeWindowPreflight(
+  connectionId: string,
+  connection?: Record<string, unknown>
+): Promise<QuotaInfo | null> {
+  const freeWindowExhausted = checkFreeWindowExhausted(
+    connectionId,
+    connection,
+    connection?.requestedModel
+  );
+  return freeWindowExhausted ?? fetchOpenrouterQuota(connectionId, connection);
+}
+
+// ─── Registration ─────────────────────────────────────────────────────────────
+
+/**
+ * Register the OpenRouter quota fetcher with the preflight and monitor systems.
+ * Call this once at server startup (in chat.ts or app entry point).
+ */
+export function registerOpenrouterQuotaFetcher(): void {
+  registerQuotaFetcher("openrouter", fetchOpenrouterQuotaWithFreeWindowPreflight);
+  registerMonitorFetcher("openrouter", fetchOpenrouterQuotaWithFreeWindowPreflight);
+}
diff --git a/open-sse/services/usage.ts b/open-sse/services/usage.ts
index 1083748e9b..4e24f00322 100644
--- a/open-sse/services/usage.ts
+++ b/open-sse/services/usage.ts
@@ -7,6 +7,7 @@ import { getDbInstance } from "@/lib/db/core";
 import { fetchBailianQuota, type BailianTripleWindowQuota } from "./bailianQuotaFetcher.ts";
 import { fetchDeepseekQuota, type DeepseekQuota } from "./deepseekQuotaFetcher.ts";
 import { fetchOpencodeQuota, type OpencodeTripleWindowQuota } from "./opencodeQuotaFetcher.ts";
+import { getOpenrouterUsage } from "./usage/openrouter.ts";
 import { getOllamaCloudUsage, getOpenCodeGoUsage } from "./opencodeOllamaUsage.ts";
 import { getCodeBuddyCnUsage } from "./usage/codebuddy-cn.ts";
 import {
@@ -539,6 +540,7 @@ export const USAGE_FETCHER_PROVIDERS = [
   "vertex",
   "vertex-partner",
   "codebuddy-cn",
+  "openrouter",
 ] as const;
 
 export type UsageFetcherProvider = (typeof USAGE_FETCHER_PROVIDERS)[number];
@@ -612,6 +614,8 @@ export async function getUsageForProvider(
       return await getNanoGptUsage(apiKey || "");
     case "deepseek":
       return await getDeepseekUsage(id || "", apiKey || "");
+    case "openrouter":
+      return await getOpenrouterUsage(id || "", apiKey || "", providerSpecificData);
     case "opencode":
     case "opencode-zen":
       return await getOpencodeUsage(id || "", apiKey || "");
diff --git a/open-sse/services/usage/openrouter.ts b/open-sse/services/usage/openrouter.ts
new file mode 100644
index 0000000000..815dd083ef
--- /dev/null
+++ b/open-sse/services/usage/openrouter.ts
@@ -0,0 +1,90 @@
+/**
+ * usage/openrouter.ts — OpenRouter usage-dashboard builder (#6842)
+ *
+ * Extracted as a leaf module (not inlined in usage.ts) so the god-file stays
+ * flat: this owns turning an OpenrouterQuota into the `UsageQuota` shape the
+ * Dashboard → Usage page renders, mirroring getDeepseekUsage's pattern.
+ */
+
+import { fetchOpenrouterQuota, type OpenrouterQuota } from "../openrouterQuotaFetcher.ts";
+import { getFreeWindowStatus, resolveAccountKey } from "../openrouterFreeWindow.ts";
+import { type UsageQuota } from "./quota.ts";
+
+function buildCreditsQuota(quota: OpenrouterQuota): UsageQuota | null {
+  if (quota.limit === null && quota.creditBalance === null) return null;
+  return {
+    used: quota.limit !== null ? quota.limit - (quota.limitRemaining ?? quota.limit) : 0,
+    total: quota.limit ?? 0,
+    remaining: quota.creditBalance ?? undefined,
+    remainingPercentage: quota.limit !== null ? Math.round((1 - quota.percentUsed) * 100) : 100,
+    resetAt: quota.resetAt ?? null,
+    unlimited: quota.limit === null,
+    currency: "USD",
+  };
+}
+
+function buildFreeWindowQuota(connectionId: string, connection?: Record<string, unknown>) {
+  const accountKey = resolveAccountKey(connectionId, connection);
+  const status = getFreeWindowStatus(accountKey);
+  const dailyQuota: UsageQuota = {
+    used: status.dailyUsed,
+    total: status.dailyLimit,
+    remaining: status.dailyRemaining,
+    remainingPercentage: Math.round((status.dailyRemaining / status.dailyLimit) * 100),
+    resetAt: status.dailyResetAt,
+    unlimited: false,
+    displayName: "Free-tier requests (daily)",
+  };
+  const rpmQuota: UsageQuota = {
+    used: status.rpmUsed,
+    total: status.rpmLimit,
+    remaining: status.rpmRemaining,
+    remainingPercentage: Math.round((status.rpmRemaining / status.rpmLimit) * 100),
+    resetAt: null,
+    unlimited: false,
+    displayName: "Free-tier requests (per minute)",
+  };
+  return { dailyQuota, rpmQuota };
+}
+
+/**
+ * OpenRouter Usage — merges the /key + /credits polling fetcher with the
+ * locally-tracked `:free`-variant request window into one usage payload.
+ */
+export async function getOpenrouterUsage(
+  connectionId: string,
+  apiKey: string,
+  providerSpecificData?: Record<string, unknown> | null
+) {
+  if (!apiKey) {
+    return { message: "OpenRouter API key not available. Add a key to view usage." };
+  }
+
+  const connection = { apiKey, providerSpecificData: providerSpecificData ?? {} };
+  const quota = (await fetchOpenrouterQuota(connectionId, connection)) as OpenrouterQuota | null;
+
+  const quotas: Record<string, UsageQuota> = {};
+  const { dailyQuota, rpmQuota } = buildFreeWindowQuota(connectionId, connection);
+  quotas.free_daily = dailyQuota;
+  quotas.free_rpm = rpmQuota;
+
+  if (!quota) {
+    return {
+      plan: "OpenRouter (usage endpoint unreachable)",
+      quotas,
+      message: "OpenRouter connected. Balance/credit-cap data temporarily unavailable.",
+    };
+  }
+
+  const creditsQuota = buildCreditsQuota(quota);
+  if (creditsQuota) quotas.credits = creditsQuota;
+
+  return {
+    plan: quota.isFreeTier ? "OpenRouter (Free Tier)" : "OpenRouter",
+    quotas,
+    isFreeTier: quota.isFreeTier,
+    usageDaily: quota.usageDaily,
+    usageWeekly: quota.usageWeekly,
+    usageMonthly: quota.usageMonthly,
+  };
+}
diff --git a/src/sse/handlers/chat.ts b/src/sse/handlers/chat.ts
index 0bee6caa42..32b5163fef 100644
--- a/src/sse/handlers/chat.ts
+++ b/src/sse/handlers/chat.ts
@@ -124,6 +124,7 @@ import {
 import { registerBailianCodingPlanQuotaFetcher } from "@omniroute/open-sse/services/bailianQuotaFetcher.ts";
 import { registerCrofUsageFetcher } from "@omniroute/open-sse/services/crofUsageFetcher.ts";
 import { registerDeepseekQuotaFetcher } from "@omniroute/open-sse/services/deepseekQuotaFetcher.ts";
+import { registerOpenrouterQuotaFetcher } from "@omniroute/open-sse/services/openrouterQuotaFetcher.ts";
 import { registerOpencodeQuotaFetcher } from "@omniroute/open-sse/services/opencodeQuotaFetcher.ts";
 import { registerGenericQuotaFetchers } from "@omniroute/open-sse/services/genericQuotaFetcher.ts";
 import {
@@ -145,10 +146,10 @@ registerBailianCodingPlanQuotaFetcher();
 // Surfaces usable_requests + credits in the monitor and only blocks (preflight
 // opt-in) when the active bucket reaches zero.
 registerCrofUsageFetcher();
-
 // Register DeepSeek balance quota fetcher.
 // Hooks into quotaPreflight + quotaMonitor so combos can switch accounts before balance is exhausted.
 registerDeepseekQuotaFetcher();
+registerOpenrouterQuotaFetcher();
 
 // Register OpenCode quota fetcher (opencode-go / opencode / opencode-zen).
 // Surfaces the $12/5h, $30/wk, $60/mo windows in the limits page and enables
diff --git a/src/sse/services/auth.ts b/src/sse/services/auth.ts
index aff185ef83..3d7bff3d1e 100644
--- a/src/sse/services/auth.ts
+++ b/src/sse/services/auth.ts
@@ -1483,8 +1483,7 @@ export async function getProviderCredentials(
       { fallbackStrategy?: string; stickyRoundRobinLimit?: number }
     >;
     const providerOverride = providerStrategyOverrides[resolvedId] || {};
-    const strategy =
-      providerOverride.fallbackStrategy || settings.fallbackStrategy || "fill-first";
+    const strategy = providerOverride.fallbackStrategy || settings.fallbackStrategy || "fill-first";
 
     let connection;
     const affinityConnection = await selectSessionAffinityConnection(
@@ -1834,8 +1833,10 @@ export async function getProviderCredentialsWithQuotaPreflight(
       }
       return defaultThresholdPercent;
     };
+    // #6842: openrouter also needs requestedModel, for the :free-window check.
+    const modelAwarePreflight = provider === "codex" || provider === "openrouter";
     const preflightCredentials =
-      requestedModel && provider === "codex" ? { ...credentials, requestedModel } : credentials;
+      requestedModel && modelAwarePreflight ? { ...credentials, requestedModel } : credentials;
     const preflight = await preflightQuota(provider, connectionId, preflightCredentials, {
       resolveMinRemainingPercent,
       resolveWarnRemainingPercent: () => warnThresholdPercent,
diff --git a/tests/unit/openrouter-free-window-wiring-6842.test.ts b/tests/unit/openrouter-free-window-wiring-6842.test.ts
new file mode 100644
index 0000000000..ec357da6a2
--- /dev/null
+++ b/tests/unit/openrouter-free-window-wiring-6842.test.ts
@@ -0,0 +1,223 @@
+/**
+ * #6842 follow-up — wire the OpenRouter `:free`-window counter into the
+ * actual request pipeline (dispatch + combo preflight skip).
+ *
+ * The counter itself (open-sse/services/openrouterFreeWindow.ts) shipped in
+ * #6842 but was never called anywhere outside the dashboard usage reader —
+ * combos kept spending guaranteed-429 requests on exhausted `:free` targets.
+ * This file proves the two wiring chokepoints:
+ *
+ *   1. RECORD — open-sse/executors/base.ts dispatches a `:free`-variant
+ *      OpenRouter request -> recordFreeWindowAttempt() fires, and a 429
+ *      response carrying X-RateLimit-* headers self-corrects the local
+ *      counter via correctFromRateLimitHeaders().
+ *   2. ENFORCE — open-sse/services/openrouterQuotaFetcher.ts's
+ *      fetchOpenrouterQuotaWithFreeWindowPreflight() (the function actually
+ *      registered with quotaPreflight.ts / quotaMonitor.ts — the same
+ *      chokepoint the PR's own /key+/credits check uses) returns
+ *      limitReached:true for an exhausted `:free` model WITHOUT making any
+ *      network call — proving combos skip the target instead of dispatching
+ *      a guaranteed-429 request. It wraps the unchanged fetchOpenrouterQuota()
+ *      so the free-window short-circuit doesn't add branching to that
+ *      function's own (already tight) cyclomatic-complexity budget.
+ */
+import test from "node:test";
+import assert from "node:assert/strict";
+
+import { DefaultExecutor } from "../../open-sse/executors/default.ts";
+import {
+  clearFreeWindowState,
+  getFreeWindowStatus,
+  resolveAccountKey,
+} from "../../open-sse/services/openrouterFreeWindow.ts";
+import {
+  fetchOpenrouterQuotaWithFreeWindowPreflight,
+  invalidateOpenrouterQuotaCache,
+} from "../../open-sse/services/openrouterQuotaFetcher.ts";
+
+const originalFetch = globalThis.fetch;
+
+test.afterEach(() => {
+  globalThis.fetch = originalFetch;
+  clearFreeWindowState();
+});
+
+// ─── 1. RECORD: dispatching a `:free` request increments the window ───────
+
+test("DefaultExecutor(openrouter) records a free-window attempt when dispatching a :free model", async () => {
+  const executor = new DefaultExecutor("openrouter");
+  const connectionId = `openrouter-record-${Date.now()}`;
+  let fetchCalls = 0;
+
+  globalThis.fetch = async () => {
+    fetchCalls += 1;
+    return new Response(JSON.stringify({ choices: [] }), {
+      status: 200,
+      headers: { "Content-Type": "application/json" },
+    });
+  };
+
+  const accountKey = resolveAccountKey(connectionId, {});
+  assert.equal(getFreeWindowStatus(accountKey).dailyUsed, 0, "precondition: window starts empty");
+
+  await executor.execute({
+    model: "x-ai/grok-4-fast:free",
+    body: { model: "x-ai/grok-4-fast:free", messages: [{ role: "user", content: "hi" }] },
+    stream: false,
+    credentials: { apiKey: "or-test-key", connectionId },
+  });
+
+  assert.equal(fetchCalls, 1, "sanity: request was actually dispatched");
+  assert.equal(
+    getFreeWindowStatus(accountKey).dailyUsed,
+    1,
+    "recordFreeWindowAttempt should have incremented the daily counter"
+  );
+});
+
+test("DefaultExecutor(openrouter) does NOT record a free-window attempt for a non-:free model", async () => {
+  const executor = new DefaultExecutor("openrouter");
+  const connectionId = `openrouter-nonfree-${Date.now()}`;
+
+  globalThis.fetch = async () =>
+    new Response(JSON.stringify({ choices: [] }), {
+      status: 200,
+      headers: { "Content-Type": "application/json" },
+    });
+
+  const accountKey = resolveAccountKey(connectionId, {});
+
+  await executor.execute({
+    model: "x-ai/grok-4-fast",
+    body: { model: "x-ai/grok-4-fast", messages: [{ role: "user", content: "hi" }] },
+    stream: false,
+    credentials: { apiKey: "or-test-key", connectionId },
+  });
+
+  assert.equal(
+    getFreeWindowStatus(accountKey).dailyUsed,
+    0,
+    "paid (non-:free) OpenRouter models must not touch the free-window counter"
+  );
+});
+
+test("DefaultExecutor(openrouter) self-corrects the free window from X-RateLimit-* response headers on 429", async () => {
+  const executor = new DefaultExecutor("openrouter");
+  const connectionId = `openrouter-correct-${Date.now()}`;
+
+  globalThis.fetch = async () =>
+    new Response(JSON.stringify({ error: "rate limited" }), {
+      status: 429,
+      headers: {
+        "Content-Type": "application/json",
+        "x-ratelimit-limit": "50",
+        "x-ratelimit-remaining": "0",
+        "x-ratelimit-reset": String(Math.floor(Date.now() / 1000) + 3600),
+      },
+    });
+
+  const accountKey = resolveAccountKey(connectionId, {});
+
+  await executor.execute({
+    model: "x-ai/grok-4-fast:free",
+    body: { model: "x-ai/grok-4-fast:free", messages: [{ role: "user", content: "hi" }] },
+    stream: false,
+    credentials: { apiKey: "or-test-key", connectionId },
+    skipUpstreamRetry: true,
+  });
+
+  const status = getFreeWindowStatus(accountKey);
+  assert.equal(status.dailyRemaining, 0, "server-reported remaining should override the local count");
+  assert.equal(status.dailyLimit, 50, "server-reported limit should be adopted");
+});
+
+// ─── 2. ENFORCE: exhausted free window short-circuits the quota preflight ─
+
+test("fetchOpenrouterQuotaWithFreeWindowPreflight returns limitReached for an exhausted :free model WITHOUT calling fetch (RED without wiring)", async () => {
+  const connectionId = `openrouter-enforce-${Date.now()}`;
+  const accountKey = resolveAccountKey(connectionId, {});
+
+  // Exhaust the daily window (default cap: 50/day at $0 purchased-tier).
+  for (let i = 0; i < 50; i++) {
+    const { recordFreeWindowAttempt } = await import(
+      "../../open-sse/services/openrouterFreeWindow.ts"
+    );
+    recordFreeWindowAttempt(accountKey);
+  }
+  assert.equal(getFreeWindowStatus(accountKey).dailyRemaining, 0, "precondition: window exhausted");
+
+  let fetchCalls = 0;
+  globalThis.fetch = async () => {
+    fetchCalls += 1;
+    return new Response(JSON.stringify({ data: {} }), { status: 200 });
+  };
+
+  const quota = await fetchOpenrouterQuotaWithFreeWindowPreflight(connectionId, {
+    apiKey: "or-test-key",
+    requestedModel: "x-ai/grok-4-fast:free",
+  });
+
+  assert.equal(fetchCalls, 0, "an exhausted free window must short-circuit BEFORE any /key or /credits call");
+  assert.ok(quota, "quota should be a limitReached result, not null");
+  assert.equal(quota!.limitReached, true, "combo preflight must see limitReached to skip this target");
+  invalidateOpenrouterQuotaCache(connectionId);
+});
+
+test("fetchOpenrouterQuotaWithFreeWindowPreflight proceeds to the normal /key+/credits fetch when the free window is NOT exhausted", async () => {
+  const connectionId = `openrouter-enforce-ok-${Date.now()}`;
+
+  let fetchCalls = 0;
+  globalThis.fetch = async (url) => {
+    fetchCalls += 1;
+    if (String(url).endsWith("/key")) {
+      return new Response(
+        JSON.stringify({
+          data: { limit: null, limit_remaining: null, limit_reset: null, is_free_tier: true },
+        }),
+        { status: 200 }
+      );
+    }
+    return new Response(JSON.stringify({ data: {} }), { status: 200 });
+  };
+
+  const quota = await fetchOpenrouterQuotaWithFreeWindowPreflight(connectionId, {
+    apiKey: "or-test-key",
+    requestedModel: "x-ai/grok-4-fast:free",
+  });
+
+  assert.ok(fetchCalls > 0, "with window remaining, the fetcher must still hit /key + /credits");
+  assert.ok(quota);
+  invalidateOpenrouterQuotaCache(connectionId);
+});
+
+test("fetchOpenrouterQuotaWithFreeWindowPreflight ignores free-window state for non-:free requestedModel", async () => {
+  const connectionId = `openrouter-enforce-nonfree-${Date.now()}`;
+  const accountKey = resolveAccountKey(connectionId, {});
+  const { recordFreeWindowAttempt } = await import(
+    "../../open-sse/services/openrouterFreeWindow.ts"
+  );
+  for (let i = 0; i < 50; i++) recordFreeWindowAttempt(accountKey);
+
+  let fetchCalls = 0;
+  globalThis.fetch = async (url) => {
+    fetchCalls += 1;
+    if (String(url).endsWith("/key")) {
+      return new Response(
+        JSON.stringify({
+          data: { limit: null, limit_remaining: null, limit_reset: null, is_free_tier: false },
+        }),
+        { status: 200 }
+      );
+    }
+    return new Response(JSON.stringify({ data: {} }), { status: 200 });
+  };
+
+  const quota = await fetchOpenrouterQuotaWithFreeWindowPreflight(connectionId, {
+    apiKey: "or-test-key",
+    requestedModel: "x-ai/grok-4-fast", // paid variant — exhausted free window is irrelevant
+  });
+
+  assert.ok(fetchCalls > 0, "a paid model must never be short-circuited by the free-window counter");
+  assert.ok(quota);
+  invalidateOpenrouterQuotaCache(connectionId);
+});
diff --git a/tests/unit/openrouter-quota-6842.test.ts b/tests/unit/openrouter-quota-6842.test.ts
new file mode 100644
index 0000000000..20b4818f40
--- /dev/null
+++ b/tests/unit/openrouter-quota-6842.test.ts
@@ -0,0 +1,345 @@
+/**
+ * #6842 — OpenRouter quota tracking (TDD).
+ *
+ * Covers the three scoped behaviors:
+ *   1. /api/v1/key + /api/v1/credits endpoint parsing (incl. null fields, BYOK).
+ *   2. Free-window local counter: UTC-day rollover + 20 RPM rolling window +
+ *      X-RateLimit-* header correction.
+ *   3. 402 -> credit-exhausted classification (connection-scope lock, not the
+ *      whole-provider circuit breaker).
+ */
+import test from "node:test";
+import assert from "node:assert/strict";
+
+import {
+  fetchOpenrouterQuota,
+  invalidateOpenrouterQuotaCache,
+  parseOpenrouterKeyResponse,
+  parseOpenrouterCreditsResponse,
+  registerOpenrouterQuotaFetcher,
+} from "../../open-sse/services/openrouterQuotaFetcher.ts";
+import {
+  clearFreeWindowState,
+  correctFromRateLimitHeaders,
+  getFreeWindowStatus,
+  recordFreeWindowAttempt,
+  resolveAccountKey,
+  setPurchasedTier,
+} from "../../open-sse/services/openrouterFreeWindow.ts";
+import { getProviderErrorRuleMatch } from "../../open-sse/config/providerErrorRules.ts";
+
+const originalFetch = globalThis.fetch;
+
+test.afterEach(() => {
+  globalThis.fetch = originalFetch;
+  clearFreeWindowState();
+});
+
+// ─── 1. Endpoint parsing ──────────────────────────────────────────────────
+
+test("parseOpenrouterKeyResponse parses a full /api/v1/key response", () => {
+  const parsed = parseOpenrouterKeyResponse({
+    data: {
+      limit: 100,
+      limit_remaining: 42.5,
+      limit_reset: null,
+      is_free_tier: false,
+      usage: 57.5,
+      usage_daily: 1.2,
+      usage_weekly: 8,
+      usage_monthly: 20,
+      byok_usage: 3.1,
+      include_byok_in_limit: true,
+    },
+  });
+  assert.ok(parsed);
+  assert.equal(parsed.limit, 100);
+  assert.equal(parsed.limitRemaining, 42.5);
+  assert.equal(parsed.limitReset, null);
+  assert.equal(parsed.isFreeTier, false);
+  assert.equal(parsed.usageDaily, 1.2);
+  assert.equal(parsed.byokUsage, 3.1);
+  assert.equal(parsed.includeByokInLimit, true);
+});
+
+test("parseOpenrouterKeyResponse treats null limit/limit_remaining as unlimited", () => {
+  const parsed = parseOpenrouterKeyResponse({
+    data: {
+      limit: null,
+      limit_remaining: null,
+      limit_reset: null,
+      is_free_tier: true,
+      usage: 0,
+      usage_daily: 0,
+      usage_weekly: 0,
+      usage_monthly: 0,
+    },
+  });
+  assert.ok(parsed);
+  assert.equal(parsed.limit, null);
+  assert.equal(parsed.limitRemaining, null);
+  assert.equal(parsed.isFreeTier, true);
+  // BYOK fields absent entirely — must not throw, must default to null/false.
+  assert.equal(parsed.byokUsage, null);
+  assert.equal(parsed.includeByokInLimit, false);
+});
+
+test("parseOpenrouterKeyResponse returns null for an unrecognizable payload", () => {
+  assert.equal(parseOpenrouterKeyResponse({}), null);
+  assert.equal(parseOpenrouterKeyResponse(null), null);
+  assert.equal(parseOpenrouterKeyResponse("not json"), null);
+});
+
+test("parseOpenrouterCreditsResponse parses total_credits/total_usage", () => {
+  const parsed = parseOpenrouterCreditsResponse({
+    data: { total_credits: 50, total_usage: 12.5 },
+  });
+  assert.equal(parsed.totalCredits, 50);
+  assert.equal(parsed.totalUsage, 12.5);
+});
+
+test("parseOpenrouterCreditsResponse defaults to nulls on missing data", () => {
+  const parsed = parseOpenrouterCreditsResponse({});
+  assert.equal(parsed.totalCredits, null);
+  assert.equal(parsed.totalUsage, null);
+});
+
+test("fetchOpenrouterQuota returns null when no API key exists", async () => {
+  const quota = await fetchOpenrouterQuota(`missing-${Date.now()}`);
+  assert.equal(quota, null);
+});
+
+test("fetchOpenrouterQuota merges /key + /credits into one quota", async () => {
+  const connectionId = `openrouter-merge-${Date.now()}`;
+  const calls: string[] = [];
+
+  globalThis.fetch = async (url) => {
+    const href = String(url);
+    calls.push(href);
+    if (href.endsWith("/key")) {
+      return new Response(
+        JSON.stringify({
+          data: {
+            limit: 100,
+            limit_remaining: 10,
+            limit_reset: null,
+            is_free_tier: false,
+            usage: 90,
+            usage_daily: 5,
+            usage_weekly: 20,
+            usage_monthly: 90,
+          },
+        }),
+        { status: 200 }
+      );
+    }
+    return new Response(JSON.stringify({ data: { total_credits: 100, total_usage: 90 } }), {
+      status: 200,
+    });
+  };
+
+  const quota = await fetchOpenrouterQuota(connectionId, { apiKey: "test-key" });
+  assert.ok(quota);
+  assert.equal(calls.length, 2);
+  assert.equal((quota as { limitReached?: boolean }).limitReached, false);
+  assert.equal(quota.percentUsed, 0.9);
+  invalidateOpenrouterQuotaCache(connectionId);
+});
+
+test("fetchOpenrouterQuota marks limitReached when limit_remaining <= 0", async () => {
+  const connectionId = `openrouter-exhausted-${Date.now()}`;
+
+  globalThis.fetch = async (url) => {
+    if (String(url).endsWith("/key")) {
+      return new Response(
+        JSON.stringify({
+          data: { limit: 10, limit_remaining: 0, limit_reset: null, is_free_tier: true },
+        }),
+        { status: 200 }
+      );
+    }
+    return new Response(JSON.stringify({ data: {} }), { status: 200 });
+  };
+
+  const quota = await fetchOpenrouterQuota(connectionId, { apiKey: "test-key" });
+  assert.ok(quota);
+  assert.equal(quota.limitReached, true);
+  invalidateOpenrouterQuotaCache(connectionId);
+});
+
+test("fetchOpenrouterQuota returns null on 401 (invalid token)", async () => {
+  const connectionId = `openrouter-401-${Date.now()}`;
+  globalThis.fetch = async () => new Response(null, { status: 401 });
+  const quota = await fetchOpenrouterQuota(connectionId, { apiKey: "bad-key" });
+  assert.equal(quota, null);
+});
+
+test("registerOpenrouterQuotaFetcher does not throw", () => {
+  assert.doesNotThrow(() => registerOpenrouterQuotaFetcher());
+});
+
+// ─── 2. Free-window local counter ─────────────────────────────────────────
+
+test("resolveAccountKey groups multiple connections under one explicit account", () => {
+  const shared = resolveAccountKey("conn-a", {
+    providerSpecificData: { openrouterAccountKey: "team-account" },
+  });
+  const sharedAgain = resolveAccountKey("conn-b", {
+    providerSpecificData: { openrouterAccountKey: "team-account" },
+  });
+  assert.equal(shared, sharedAgain);
+
+  const unrelated = resolveAccountKey("conn-c");
+  assert.notEqual(unrelated, shared);
+});
+
+test("getFreeWindowStatus defaults to 50/day before any purchase", () => {
+  const accountKey = `acct-default-${Date.now()}`;
+  const status = getFreeWindowStatus(accountKey);
+  assert.equal(status.dailyLimit, 50);
+  assert.equal(status.dailyUsed, 0);
+  assert.equal(status.rpmLimit, 20);
+});
+
+test("setPurchasedTier unlocks the 1000/day tier", () => {
+  const accountKey = `acct-purchased-${Date.now()}`;
+  setPurchasedTier(accountKey, true);
+  const status = getFreeWindowStatus(accountKey);
+  assert.equal(status.dailyLimit, 1000);
+});
+
+test("recordFreeWindowAttempt increments the UTC-day counter", () => {
+  const accountKey = `acct-day-${Date.now()}`;
+  const now = Date.UTC(2026, 0, 15, 12, 0, 0);
+  recordFreeWindowAttempt(accountKey, now);
+  recordFreeWindowAttempt(accountKey, now + 1000);
+  const status = getFreeWindowStatus(accountKey, now + 2000);
+  assert.equal(status.dailyUsed, 2);
+  assert.equal(status.dailyRemaining, 48);
+});
+
+test("UTC-day counter rolls over at midnight", () => {
+  const accountKey = `acct-rollover-${Date.now()}`;
+  const dayOne = Date.UTC(2026, 0, 15, 23, 59, 59);
+  recordFreeWindowAttempt(accountKey, dayOne);
+  const dayOneStatus = getFreeWindowStatus(accountKey, dayOne);
+  assert.equal(dayOneStatus.dailyUsed, 1);
+
+  // Same account, next UTC day — counter must reset to 0.
+  const dayTwo = Date.UTC(2026, 0, 16, 0, 0, 1);
+  const dayTwoStatus = getFreeWindowStatus(accountKey, dayTwo);
+  assert.equal(dayTwoStatus.dailyUsed, 0);
+});
+
+test("20 RPM rolling window blocks the 21st request within 60s, frees up after", () => {
+  const accountKey = `acct-rpm-${Date.now()}`;
+  const base = Date.UTC(2026, 0, 15, 12, 0, 0);
+  for (let i = 0; i < 20; i++) {
+    recordFreeWindowAttempt(accountKey, base + i * 1000);
+  }
+  const atLimit = getFreeWindowStatus(accountKey, base + 20_000);
+  assert.equal(atLimit.rpmUsed, 20);
+  assert.equal(atLimit.rpmRemaining, 0);
+
+  // Just over 60s after the FIRST request only, that timestamp has rolled
+  // out of the window (offset by 1ms from the second request to avoid a
+  // boundary tie between the 1st and 2nd requests, which were 1000ms apart).
+  const afterRoll = getFreeWindowStatus(accountKey, base + 60_001);
+  assert.equal(afterRoll.rpmUsed, 19);
+  assert.equal(afterRoll.rpmRemaining, 1);
+});
+
+test("failed attempts count toward the daily cap", () => {
+  const accountKey = `acct-failed-${Date.now()}`;
+  const now = Date.UTC(2026, 0, 15, 12, 0, 0);
+  // The module doesn't distinguish success/failure at record time — callers
+  // record every attempt, successful or not, matching the plan's "failed
+  // attempts count toward the daily cap" requirement.
+  recordFreeWindowAttempt(accountKey, now);
+  recordFreeWindowAttempt(accountKey, now + 1);
+  const status = getFreeWindowStatus(accountKey, now + 2);
+  assert.equal(status.dailyUsed, 2);
+});
+
+test("correctFromRateLimitHeaders overrides local daily count from X-RateLimit-*", () => {
+  const accountKey = `acct-header-${Date.now()}`;
+  const now = Date.UTC(2026, 0, 15, 12, 0, 0);
+  recordFreeWindowAttempt(accountKey, now); // local thinks 1 used
+
+  const headers = new Headers({
+    "x-ratelimit-limit": "50",
+    "x-ratelimit-remaining": "3",
+    "x-ratelimit-reset": String(Math.floor((now + 3_600_000) / 1000)),
+  });
+  correctFromRateLimitHeaders(accountKey, headers, now + 500);
+
+  const status = getFreeWindowStatus(accountKey, now + 600);
+  assert.equal(status.dailyLimit, 50);
+  assert.equal(status.dailyUsed, 47); // 50 - 3 remaining, server-authoritative
+  assert.equal(status.dailyResetAt, new Date(now + 3_600_000).toISOString());
+});
+
+test("correctFromRateLimitHeaders honors Retry-After when later than X-RateLimit-Reset", () => {
+  const accountKey = `acct-retry-after-${Date.now()}`;
+  const now = Date.UTC(2026, 0, 15, 12, 0, 0);
+
+  const headers = new Headers({
+    "x-ratelimit-limit": "50",
+    "x-ratelimit-remaining": "0",
+    "x-ratelimit-reset": String(Math.floor((now + 60_000) / 1000)),
+    "retry-after": "300", // 5 minutes — later than the header reset above
+  });
+  correctFromRateLimitHeaders(accountKey, headers, now);
+
+  const status = getFreeWindowStatus(accountKey, now + 1);
+  assert.equal(status.dailyResetAt, new Date(now + 300_000).toISOString());
+});
+
+// ─── 3. 402 -> credit-exhausted classification ────────────────────────────
+
+test("openrouter 402 provider rule locks the connection, not the model", () => {
+  const match = getProviderErrorRuleMatch("openrouter", 402, {});
+  assert.ok(match);
+  assert.equal(match.reason, "quota_exhausted");
+  assert.equal(match.scope, "connection");
+  assert.ok(typeof match.cooldownMs === "number" && match.cooldownMs > 0);
+});
+
+test("openrouter 402 does not match for other status codes", () => {
+  assert.equal(getProviderErrorRuleMatch("openrouter", 429, {}), null);
+  assert.equal(getProviderErrorRuleMatch("openrouter", 500, {}), null);
+});
+
+test("checkFallbackError classifies OpenRouter 402 as quota_exhausted with a real cooldown", async () => {
+  const { checkFallbackError } = await import("../../open-sse/services/accountFallback.ts");
+  const { RateLimitReason } = await import("../../open-sse/config/constants.ts");
+
+  // errorText intentionally null/generic (no CREDITS_EXHAUSTED_SIGNALS phrase
+  // like "payment required") so this exercises the STATUS-only path — the
+  // generic `status_402` configured rule (open-sse/config/errorConfig.ts)
+  // falling through to the new provider-rule override in
+  // accountFallback.ts, not the earlier errorText-signal branch.
+  const genericResult = checkFallbackError(402, null, 0, null, "some-other-provider", null);
+  assert.equal(genericResult.shouldFallback, true);
+  assert.equal(genericResult.reason, RateLimitReason.QUOTA_EXHAUSTED);
+  // Generic default: bare cooldownMs 0 — immediate reselection of the same
+  // connection (no provider-specific rule registered for this provider).
+  assert.equal(genericResult.cooldownMs, 0);
+
+  const openrouterResult = checkFallbackError(402, null, 0, null, "openrouter", null);
+  assert.equal(openrouterResult.shouldFallback, true);
+  assert.equal(openrouterResult.reason, RateLimitReason.QUOTA_EXHAUSTED);
+  // The OpenRouter-specific rule must win over the generic status_402
+  // default and lock the connection for a real cooldown instead of 0.
+  assert.ok(
+    openrouterResult.cooldownMs > 0,
+    "expected a non-zero cooldown for a credit-exhausted 402"
+  );
+});
+
+test("checkFallbackError 402 does NOT trip the whole-provider circuit breaker set", async () => {
+  // Per CLAUDE.md's resilience-layer contract, only 408/500/502/503/504 trip
+  // the provider breaker. 402 must never be in that set, for any provider.
+  const PROVIDER_BREAKER_STATUSES = new Set([408, 500, 502, 503, 504]);
+  assert.equal(PROVIDER_BREAKER_STATUSES.has(402), false);
+});