diff --git a/server-dev/deploy.sh b/server-dev/deploy.sh
index 08c1ccb..117c598 100755
--- a/server-dev/deploy.sh
+++ b/server-dev/deploy.sh
@@ -63,7 +63,7 @@ if [[ ! -f "$ENV_FILE" ]]; then
 fi
 
 log "[Phase 4] start-all-server.sh (빌드 포함)"
-"$WORK_TREE/server-dev/start-all-server.sh"
+"$WORK_TREE/server-dev/start-all-server.sh" --skip-legacy-stop
 
 log "======================================================"
 log "  배포 완료 — $COMMIT"
diff --git a/server-dev/start-all-server.sh b/server-dev/start-all-server.sh
index 132f268..af7bd6a 100755
--- a/server-dev/start-all-server.sh
+++ b/server-dev/start-all-server.sh
@@ -104,10 +104,18 @@ ensure_host_ai() {
     return 0
   fi
   local ai_script="${LEGACY_STACK_DIR}/scripts/exdev/start-host-ai.sh"
-  [[ -x "${ai_script}" ]] || [[ -f "${ai_script}" ]] || fail "호스트 AI 스크립트 없음: ${ai_script}"
-  log "호스트 AI 기동..."
-  bash "${ai_script}" || fail "start-host-ai.sh 실패"
-  wait_http "http://127.0.0.1:${LLM_PORT}/v1/models" "호스트 LLM" 120 || true
+  if [[ -x "${ai_script}" ]] || [[ -f "${ai_script}" ]]; then
+    log "호스트 AI 기동..."
+    bash "${ai_script}" || log "⚠️  start-host-ai.sh 일부 실패 — 기존 프로세스 확인"
+  fi
+  if host_ai_ready; then
+    return 0
+  fi
+  if curl -sf --max-time 3 "http://127.0.0.1:${LLM_PORT}/v1/models" >/dev/null 2>&1; then
+    log "⚠️  LLM만 응답 (Embed/Rerank 미기동) — degraded 모드로 계속"
+    return 0
+  fi
+  fail "호스트 AI 미응답 (:${LLM_PORT})"
 }
 
 stop_legacy_stack() {