diff --git a/keys.mjs b/keys.mjs
index c597fe7..7d6d3f9 100644
--- a/keys.mjs
+++ b/keys.mjs
@@ -371,6 +371,36 @@ export function getCacheStats() {
   return { entries: total, totalHits, sizeBytes };
 }
 
+// ── Singleflight stampede protection ──
+
+// In-memory singleflight Map: hash → { promise, requesters }
+// Deduplicates concurrent identical cache-miss flows so only one upstream call runs.
+// Per ADR 0005 / spec D4: in-process scope only (single Node process per host).
+const inflightMap = new Map();
+
+export function singleflight(hash, fn) {
+  const existing = inflightMap.get(hash);
+  if (existing) {
+    existing.requesters++;
+    return existing.promise;
+  }
+  // Wrap fn() in Promise.resolve().then() so synchronous throws don't escape.
+  const promise = Promise.resolve().then(fn).finally(() => {
+    inflightMap.delete(hash);
+  });
+  inflightMap.set(hash, { promise, requesters: 1 });
+  return promise;
+}
+
+export function getInflightStats() {
+  let totalRequesters = 0;
+  for (const entry of inflightMap.values()) totalRequesters += entry.requesters;
+  return {
+    inflight: inflightMap.size,
+    requesters: totalRequesters,
+  };
+}
+
 // Find a key by id or name (returns { id, name } or null)
 export function findKey(idOrName) {
   const d = getDb();
diff --git a/server.mjs b/server.mjs
index 72c936d..11b0aae 100644
--- a/server.mjs
+++ b/server.mjs
@@ -34,7 +34,7 @@ import { readFileSync, accessSync, existsSync, constants } from "node:fs";
 import { fileURLToPath } from "node:url";
 import { dirname, join } from "node:path";
 import { homedir } from "node:os";
-import { validateKey, recordUsage, getUsageByKey, getUsageTimeline, getRecentUsage, createKey, listKeys, revokeKey, closeDb, checkQuota, updateKeyQuota, getKeyQuota, findKey, cacheHash, getCachedResponse, setCachedResponse, clearCache, getCacheStats, hasCacheControl } from "./keys.mjs";
+import { validateKey, recordUsage, getUsageByKey, getUsageTimeline, getRecentUsage, createKey, listKeys, revokeKey, closeDb, checkQuota, updateKeyQuota, getKeyQuota, findKey, cacheHash, getCachedResponse, setCachedResponse, clearCache, getCacheStats, hasCacheControl, singleflight, getInflightStats } from "./keys.mjs";
 
 const __dirname = dirname(fileURLToPath(import.meta.url));
 const _pkg = JSON.parse(readFileSync(join(__dirname, "package.json"), "utf8"));
@@ -590,6 +590,7 @@ function startHeartbeat(res, intervalMs, sessionId) {
 // ── Call claude CLI (real streaming) ─────────────────────────────────────
 // Pipes stdout from the claude process directly to SSE chunks as they arrive.
 // Each data chunk becomes a proper SSE event with delta content in real time.
+// TODO(cache-singleflight-stream): streaming-path singleflight is out of scope for v3.13.0; see spec D4 streaming caveat.
 function callClaudeStreaming(model, messages, conversationId, res, authInfo = {}) {
   const id = `chatcmpl-${randomUUID()}`;
   const created = Math.floor(Date.now() / 1000);
@@ -1265,14 +1266,45 @@ async function handleChatCompletions(req, res) {
 
   const t0Usage = Date.now();
   const promptChars = messages.reduce((a, m) => a + (typeof m.content === "string" ? m.content.length : JSON.stringify(m.content).length), 0);
+
+  // Non-streaming path with stampede protection: wrap the upstream call in singleflight
+  // when cache is enabled and a hash is present. Concurrent identical requests share
+  // one upstream spawn; followers receive the same promise. Streaming-path dedup is
+  // explicitly out of scope (see TODO comment above callClaudeStreaming).
+  if (CACHE_TTL > 0 && req._cacheHash) {
+    try {
+      const content = await singleflight(req._cacheHash, async () => {
+        // Re-check cache inside the singleflight: a follower that enters before the
+        // leader finishes will wait on the shared promise (not reach here), but a
+        // request that races in just after the previous singleflight cleared the map
+        // will re-read the freshly-populated cache entry here rather than spawning.
+        const recheck = getCachedResponse(req._cacheHash, CACHE_TTL);
+        if (recheck) return recheck.response;
+        const c = await callClaude(model, messages, conversationId);
+        try { setCachedResponse(req._cacheHash, model, c); } catch (e) { logEvent("error", "cache_write_failed", { error: e.message }); }
+        return c;
+      });
+      const id = `chatcmpl-${randomUUID()}`;
+      completionResponse(res, id, model, content);
+      try { recordUsage({ keyId: req._authKeyId, keyName: req._authKeyName, model, promptChars, responseChars: content.length, elapsedMs: Date.now() - t0Usage, success: true }); } catch (e) { logEvent("error", "usage_record_failed", { error: e.message }); }
+      return;
+    } catch (err) {
+      try { recordUsage({ keyId: req._authKeyId, keyName: req._authKeyName, model, promptChars, responseChars: 0, elapsedMs: Date.now() - t0Usage, success: false }); } catch (e) { logEvent("error", "usage_record_failed", { error: e.message }); }
+      console.error(`[proxy] error: ${err.message}`);
+      if (res.headersSent || res.writableEnded || res.destroyed) {
+        try { res.end(); } catch {}
+        return;
+      }
+      const safeMessage = (err.message || "Internal error").replace(/\/[\w/.\-]+/g, "[path]");
+      return jsonResponse(res, 500, { error: { message: safeMessage, type: "proxy_error" } });
+    }
+  }
+
+  // Fallback: cache disabled (CACHE_TTL=0) or no _cacheHash — original path untouched.
   try {
     const content = await callClaude(model, messages, conversationId);
     const id = `chatcmpl-${randomUUID()}`;
     completionResponse(res, id, model, content);
-    // Write to cache
-    if (CACHE_TTL > 0 && req._cacheHash) {
-      try { setCachedResponse(req._cacheHash, model, content); } catch (e) { logEvent("error", "cache_write_failed", { error: e.message }); }
-    }
     try { recordUsage({ keyId: req._authKeyId, keyName: req._authKeyName, model, promptChars, responseChars: content.length, elapsedMs: Date.now() - t0Usage, success: true }); } catch (e) { logEvent("error", "usage_record_failed", { error: e.message }); }
   } catch (err) {
     try { recordUsage({ keyId: req._authKeyId, keyName: req._authKeyName, model, promptChars, responseChars: 0, elapsedMs: Date.now() - t0Usage, success: false }); } catch (e) { logEvent("error", "usage_record_failed", { error: e.message }); }
@@ -1551,10 +1583,10 @@ const server = createServer(async (req, res) => {
     });
   }
 
-  // GET /cache/stats — cache statistics
+  // GET /cache/stats — cache statistics (entries, hits, size, inflight singleflight count)
   if (pathname === "/cache/stats" && req.method === "GET") {
     if (!isAdmin) return jsonResponse(res, 403, { error: "Admin access required" });
-    return jsonResponse(res, 200, getCacheStats());
+    return jsonResponse(res, 200, { ...getCacheStats(), ...getInflightStats() });
   }
 
   // DELETE /cache — clear cache
diff --git a/test-features.mjs b/test-features.mjs
index 5eccf93..63be8dc 100644
--- a/test-features.mjs
+++ b/test-features.mjs
@@ -3,7 +3,7 @@
  * Integration test for Quota + Cache features.
  * Tests database layer functions directly — no server needed.
  */
-import { getDb, createKey, listKeys, validateKey, recordUsage, checkQuota, updateKeyQuota, getKeyQuota, findKey, cacheHash, getCachedResponse, setCachedResponse, clearCache, getCacheStats, closeDb, hasCacheControl } from "./keys.mjs";
+import { getDb, createKey, listKeys, validateKey, recordUsage, checkQuota, updateKeyQuota, getKeyQuota, findKey, cacheHash, getCachedResponse, setCachedResponse, clearCache, getCacheStats, closeDb, hasCacheControl, singleflight, getInflightStats } from "./keys.mjs";
 import { createHash } from "node:crypto";
 import { strict as assert } from "node:assert";
 import { unlinkSync } from "node:fs";
@@ -354,6 +354,106 @@ test("D3: chunked replay uses Array.from — multibyte codepoints stay intact",
   }
 });
 
+// ── PR-B Singleflight tests (async) ──
+async function asyncTest(name, fn) {
+  try {
+    await fn();
+    passed++;
+    console.log(`  ✓ ${name}`);
+  } catch (e) {
+    failed++;
+    console.log(`  ✗ ${name}: ${e.message}`);
+  }
+}
+
+async function runSingleflightTests() {
+  console.log("\nPR-B Singleflight:");
+
+  // 1. Basic dedup: 10 concurrent calls with same hash execute fn only once.
+  await asyncTest("basic dedup: 10 concurrent callers execute fn only once", async () => {
+    let callCount = 0;
+    const fn = () => new Promise(resolve => {
+      callCount++;
+      setTimeout(() => resolve("result-A"), 20);
+    });
+    const results = await Promise.all(Array.from({ length: 10 }, () => singleflight("sf-dedup-1", fn)));
+    assert.equal(callCount, 1, `fn called ${callCount} times, expected 1`);
+    assert.ok(results.every(r => r === "result-A"), "all 10 callers should receive the same return value");
+  });
+
+  // 2. Failure fan-out: all followers reject when leader rejects.
+  await asyncTest("failure fan-out: all followers reject with leader error", async () => {
+    let callCount = 0;
+    const fn = () => new Promise((_, reject) => {
+      callCount++;
+      setTimeout(() => reject(new Error("upstream-fail")), 20);
+    });
+    const promises = Array.from({ length: 10 }, () => singleflight("sf-fail-1", fn));
+    const results = await Promise.allSettled(promises);
+    assert.equal(callCount, 1, `fn called ${callCount} times, expected 1`);
+    assert.ok(results.every(r => r.status === "rejected"), "all 10 should be rejected");
+    assert.ok(results.every(r => r.reason?.message === "upstream-fail"), "all should share the same error message");
+  });
+
+  // 3a. Map cleanup after success: inflight count returns to 0 after promise resolves.
+  await asyncTest("map cleanup after success: inflight=0 after promise settles", async () => {
+    const fn = () => new Promise(resolve => setTimeout(() => resolve("done"), 10));
+    await singleflight("sf-cleanup-ok", fn);
+    const stats = getInflightStats();
+    assert.equal(stats.inflight, 0, `expected inflight=0 after settlement, got ${stats.inflight}`);
+  });
+
+  // 3b. Map cleanup after failure: inflight count returns to 0 after promise rejects.
+  await asyncTest("map cleanup after failure: inflight=0 after promise rejects", async () => {
+    const fn = () => new Promise((_, reject) => setTimeout(() => reject(new Error("fail")), 10));
+    try { await singleflight("sf-cleanup-fail", fn); } catch {}
+    const stats = getInflightStats();
+    assert.equal(stats.inflight, 0, `expected inflight=0 after rejection, got ${stats.inflight}`);
+  });
+
+  // 4. Different hashes don't share: two parallel calls with distinct hashes both execute.
+  await asyncTest("different hashes do not share a singleflight entry", async () => {
+    let countA = 0;
+    let countB = 0;
+    const fnA = () => new Promise(resolve => { countA++; setTimeout(() => resolve("A"), 20); });
+    const fnB = () => new Promise(resolve => { countB++; setTimeout(() => resolve("B"), 20); });
+    const [rA, rB] = await Promise.all([singleflight("sf-hash-A", fnA), singleflight("sf-hash-B", fnB)]);
+    assert.equal(countA, 1);
+    assert.equal(countB, 1);
+    assert.equal(rA, "A");
+    assert.equal(rB, "B");
+  });
+
+  // 5. getInflightStats shape: returns { inflight: number, requesters: number }.
+  await asyncTest("getInflightStats returns correct shape", async () => {
+    // Verify shape against a settled state (inflight=0 is still the right shape).
+    const stats = getInflightStats();
+    assert.equal(typeof stats.inflight, "number", "inflight should be a number");
+    assert.equal(typeof stats.requesters, "number", "requesters should be a number");
+    // Also verify live counts: start a pending fn, check inflight>0, then resolve.
+    const { promise: blocker, resolve: resolveBlocker } = Promise.withResolvers();
+    const fn = () => blocker;
+    const p = singleflight("sf-stats-shape", fn);
+    const liveStats = getInflightStats();
+    assert.ok(liveStats.inflight >= 1, `expected inflight>=1, got ${liveStats.inflight}`);
+    resolveBlocker("ok");
+    await p;
+  });
+
+  // 6. Sequential calls don't share: singleflight is for concurrent dedup only.
+  await asyncTest("sequential calls with same hash each execute fn independently", async () => {
+    let callCount = 0;
+    const fn = () => new Promise(resolve => { callCount++; setTimeout(() => resolve(callCount), 10); });
+    const r1 = await singleflight("sf-sequential", fn);
+    const r2 = await singleflight("sf-sequential", fn);
+    assert.equal(callCount, 2, `fn should have been called twice, got ${callCount}`);
+    assert.equal(r1, 1);
+    assert.equal(r2, 2);
+  });
+}
+
+await runSingleflightTests();
+
 // ── Cleanup ──
 closeDb();