diff --git a/scripts/deploy_codex_audit_service.sh b/scripts/deploy_codex_audit_service.sh
index 869b366..c17cfef 100644
--- a/scripts/deploy_codex_audit_service.sh
+++ b/scripts/deploy_codex_audit_service.sh
@@ -207,6 +207,7 @@ Environment=CODEX_AUDIT_SERVICE_ALLOWED_REFS=${ALLOWED_REFS}
 Environment=CODEX_AUDIT_SERVICE_ALLOWED_REPOSITORY_VISIBILITIES=${ALLOWED_REPOSITORY_VISIBILITIES}
 Environment=CODEX_AUDIT_SERVICE_ALLOWED_SOURCE_REPOSITORIES=${ALLOWED_SOURCE_REPOSITORIES}
 Environment=CODEX_AUDIT_SERVICE_JOB_DIR=${JOB_DIR}
+Environment=CODEX_AUDIT_SERVICE_QUOTA_STORE=${JOB_DIR}/quota.json
 Environment=CODEX_AUDIT_SERVICE_SANDBOX=read-only
 ${audit_model_line}
 ${audit_reasoning_effort_line}
diff --git a/service/ai_gateway_service.py b/service/ai_gateway_service.py
index 0529ab3..086af94 100644
--- a/service/ai_gateway_service.py
+++ b/service/ai_gateway_service.py
@@ -756,6 +756,7 @@ def _handle_analyze(self, payload: dict[str, Any]) -> None:
 
     def _handle_execute_async(self, claims: dict[str, Any], payload: dict[str, Any]) -> None:
         """POST /v1/ai/execute/jobs — async Codex execution."""
+        started = time.time()
         req = parse_execute_request(payload)
 
         # Security: validate source_repository against allowlist
@@ -763,24 +764,27 @@ def _handle_execute_async(self, claims: dict[str, Any], payload: dict[str, Any])
         if source_repo:
             _validate_source_repo(source_repo)
             _validate_source_repo_org(claims, source_repo)
+        quota_repo = source_repo or str(claims.get("repository") or "unknown")
 
         # Quota check
         quota = get_quota_manager()
-        qr = quota.check(source_repo, "codex-cli", req.prompt)
+        qr = quota.check(quota_repo, "codex-cli", req.prompt)
         if not qr["allowed"]:
             _json_response(self, HTTPStatus.TOO_MANY_REQUESTS, {
                 "status": "error", "error": qr["reason"],
                 "remaining_usd": qr.get("remaining_usd", 0),
             })
             return
-        quota.record_execute(source_repo)
+        quota.record_execute(quota_repo)
 
         payload.setdefault("task", TASK_EXECUTE)
         job = _submit_job(claims, payload)
+        get_health_monitor().record("/v1/ai/execute/jobs", time.time() - started, True)
         _json_response(self, HTTPStatus.ACCEPTED, job)
 
     def _handle_execute_sync(self, claims: dict[str, Any], payload: dict[str, Any]) -> None:
         """POST /v1/ai/execute — sync Codex execution (backward compat)."""
+        started = time.time()
         req = parse_execute_request(payload)
         source_repo = str(payload.get("source_repository") or "")
         if source_repo:
@@ -806,6 +810,7 @@ def _handle_execute_sync(self, claims: dict[str, Any], payload: dict[str, Any])
             reasoning_effort=reasoning_effort,
             timeout=req.timeout_seconds,
         )
+        get_health_monitor().record("/v1/ai/execute", time.time() - started, result.success, result.error if not result.success else "")
         if result.success:
             _json_response(self, HTTPStatus.OK, {"status": "ok", "output": result.output})
         else:
diff --git a/service/quota.py b/service/quota.py
index b7fb90a..8089d72 100644
--- a/service/quota.py
+++ b/service/quota.py
@@ -141,6 +141,16 @@ def __init__(self):
         self._weekly_budget = DEFAULT_WEEKLY_BUDGET_USD
         self._repo_budgets: dict[str, dict[str, float]] = {}
         self._load_config()
+        self._load_records()
+
+    def _store_path(self) -> Path | None:
+        path = os.environ.get("CODEX_AUDIT_SERVICE_QUOTA_STORE", "").strip()
+        if path:
+            return Path(path)
+        job_dir = os.environ.get("CODEX_AUDIT_SERVICE_JOB_DIR", "").strip()
+        if job_dir:
+            return Path(job_dir) / "quota.json"
+        return None
 
     def _load_config(self) -> None:
         config_path = os.environ.get("CODEX_AUDIT_SERVICE_QUOTA_CONFIG", "").strip()
@@ -156,6 +166,41 @@ def _load_config(self) -> None:
             if isinstance(raw.get("repo_budgets"), dict):
                 self._repo_budgets = raw["repo_budgets"]
 
+    def _load_records(self) -> None:
+        path = self._store_path()
+        if path is None:
+            return
+        if not path.exists():
+            return
+        try:
+            raw = json.loads(path.read_text(encoding="utf-8"))
+        except (json.JSONDecodeError, OSError):
+            return
+        records = raw.get("records") if isinstance(raw, dict) else None
+        if not isinstance(records, dict):
+            return
+        self._records = {
+            repo: QuotaRecord.from_dict(item)
+            for repo, item in records.items()
+            if isinstance(repo, str) and isinstance(item, dict)
+        }
+
+    def _save_records_locked(self) -> None:
+        path = self._store_path()
+        if path is None:
+            return
+        path.parent.mkdir(mode=0o700, parents=True, exist_ok=True)
+        payload = json.dumps(
+            {"records": {repo: record.to_dict() for repo, record in self._records.items()}},
+            ensure_ascii=False,
+            sort_keys=True,
+        ).encode("utf-8")
+        tmp = path.with_suffix(path.suffix + ".tmp")
+        with open(tmp, "wb") as handle:
+            handle.write(payload)
+        os.chmod(tmp, 0o600)
+        os.replace(tmp, path)
+
     def _reset_if_needed(self, record: QuotaRecord) -> QuotaRecord:
         now = time.time()
         if now - record.last_reset_daily > 86400:
@@ -227,6 +272,7 @@ def record(self, repo: str, model: str, prompt: str, output: str = "") -> None:
                 record.codex_calls += 1
             record.total_cost_usd += cost
             self._records[repo] = record
+            self._save_records_locked()
 
     def record_execute(self, repo: str) -> None:
         """Record a codex exec call (flat cost)."""
@@ -239,6 +285,7 @@ def record_execute(self, repo: str) -> None:
             record.codex_calls += 1
             record.total_cost_usd += cost
             self._records[repo] = record
+            self._save_records_locked()
 
     def status(self, repo: str = "") -> dict[str, Any]:
         """Get quota status for a repo or all repos."""
diff --git a/tests/test_quota.py b/tests/test_quota.py
index bcb64a2..aa18c65 100644
--- a/tests/test_quota.py
+++ b/tests/test_quota.py
@@ -149,6 +149,17 @@ def test_get_weekly_budget_respects_repo_overrides(self) -> None:
         self.manager._repo_budgets["premium/repo"] = {"weekly": 250.0}
         self.assertEqual(self.manager.get_weekly_budget("premium/repo"), 250.0)
 
+    def test_records_persist_to_store(self) -> None:
+        with tempfile.TemporaryDirectory() as tmp:
+            store = str(Path(tmp) / "quota.json")
+            with patch.dict(os.environ, {"CODEX_AUDIT_SERVICE_QUOTA_STORE": store}):
+                first = QuotaManager()
+                first.record_execute("test/repo")
+                second = QuotaManager()
+                status = second.status("test/repo")
+        self.assertEqual(status["codex_calls"], 1)
+        self.assertGreater(status["total_cost_usd"], 0)
+
 
 class TestQuotaConfigLoading(unittest.TestCase):
     """Quota configuration from JSON file."""