diff --git a/deploy/compose/docker-compose.dev.yaml b/deploy/compose/docker-compose.dev.yaml
index 99204d8f..2ff1e6d 100644
--- a/deploy/compose/docker-compose.dev.yaml
+++ b/deploy/compose/docker-compose.dev.yaml
@@ -164,6 +164,9 @@ services:
     ports:
       - "${LLM_NIM_PORT:-8000}:8000"
     environment:
+      # NIM containers require NGC_API_KEY for startup/model download auth.
+      # Fall back to NVIDIA_API_KEY for older .env files created by the setup notebook.
+      - NGC_API_KEY=${NGC_API_KEY:-${NVIDIA_API_KEY:-}}
       - NVIDIA_API_KEY=${NVIDIA_API_KEY:-}
       - CUDA_VISIBLE_DEVICES=0,1,2,3
       - NVIDIA_VISIBLE_DEVICES=all
diff --git a/notebooks/setup/complete_setup_guide.ipynb b/notebooks/setup/complete_setup_guide.ipynb
index 6e1b78d..b1b0255 100644
--- a/notebooks/setup/complete_setup_guide.ipynb
+++ b/notebooks/setup/complete_setup_guide.ipynb
@@ -2464,6 +2464,9 @@
     "    ports:\n",
     "      - \"${{LLM_NIM_PORT:-8000}}:8000\"\n",
     "    environment:\n",
+    "      # NIM containers require NGC_API_KEY for startup/model download auth.\n",
+    "      # Fall back to NVIDIA_API_KEY for older .env files created by this notebook.\n",
+    "      - NGC_API_KEY=${{NGC_API_KEY:-${{NVIDIA_API_KEY:-}}}}\n",
     "      - NVIDIA_API_KEY=${{NVIDIA_API_KEY:-}}\n",
     "      - CUDA_VISIBLE_DEVICES={cuda_visible_devices}\n",
     "      - NVIDIA_VISIBLE_DEVICES=all\n",
@@ -2698,14 +2701,31 @@
     "            content,\n",
     "            flags=re.MULTILINE\n",
     "        )\n",
-    "        # API key optional for local\n",
+    "        # API key optional for local. NIM containers read NGC_API_KEY for\n",
+    "        # startup/model download auth; keep NVIDIA_API_KEY populated for clients\n",
+    "        # and older code paths.\n",
     "        if strategy_info.get(\"api_key\"):\n",
+    "            local_api_key = strategy_info[\"api_key\"]\n",
     "            content = re.sub(\n",
     "                r'^NVIDIA_API_KEY=.*$',\n",
-    "                f'NVIDIA_API_KEY={strategy_info[\"api_key\"]}',\n",
+    "                f'NVIDIA_API_KEY={local_api_key}',\n",
     "                content,\n",
     "                flags=re.MULTILINE\n",
     "            )\n",
+    "            if re.search(r'^NGC_API_KEY=.*$', content, flags=re.MULTILINE):\n",
+    "                content = re.sub(\n",
+    "                    r'^NGC_API_KEY=.*$',\n",
+    "                    f'NGC_API_KEY={local_api_key}',\n",
+    "                    content,\n",
+    "                    flags=re.MULTILINE\n",
+    "                )\n",
+    "            else:\n",
+    "                content = re.sub(\n",
+    "                    r'^(NVIDIA_API_KEY=.*)$',\n",
+    "                    rf'\\1\\nNGC_API_KEY={local_api_key}',\n",
+    "                    content,\n",
+    "                    flags=re.MULTILINE\n",
+    "                )\n",
     "        print(\"\\n✅ Configuration updated for LOCAL installation\")\n",
     "        print(\"   LLM_NIM_URL=http://localhost:8000/v1\")\n",
     "        print(\"   LLM_MODEL=nvidia/llama-3.3-nemotron-super-49b-v1.5\")\n",