fix(mcp): surface env-var fetch errors and authorize before existence check

getMCPUserEnvVars now throws on non-2xx so UserEnvVarsModal reports the
error instead of silently rendering the empty 'no per-user fields' state.

The per-user env-var endpoints now run the access check before the server
lookup so a non-admin cannot tell a missing server (404) apart from one
they lack access to (403), closing a server-id enumeration leak.

litellm/proxy/management_endpoints/mcp_management_endpoints.py

@@ -2248,13 +2248,13 @@ if MCP_AVAILABLE:
                 status_code=status.HTTP_400_BAD_REQUEST,
                 detail={"error": "User ID not found in token"},
             )
+        await _authorize_mcp_server_access(prisma_client, user_api_key_dict, server_id)
         server = await get_mcp_server(prisma_client, server_id)
         if server is None:
             raise HTTPException(
                 status_code=status.HTTP_404_NOT_FOUND,
                 detail={"error": f"MCP Server {server_id} not found"},
             )
-        await _authorize_mcp_server_access(prisma_client, user_api_key_dict, server_id)
         stored = await get_user_env_vars(prisma_client, user_id, server_id)
         return _compute_user_env_var_status(server=server, stored_values=stored)
 
@@ -2279,13 +2279,13 @@ if MCP_AVAILABLE:
                 status_code=status.HTTP_400_BAD_REQUEST,
                 detail={"error": "User ID not found in token"},
             )
+        await _authorize_mcp_server_access(prisma_client, user_api_key_dict, server_id)
         server = await get_mcp_server(prisma_client, server_id)
         if server is None:
             raise HTTPException(
                 status_code=status.HTTP_404_NOT_FOUND,
                 detail={"error": f"MCP Server {server_id} not found"},
             )
-        await _authorize_mcp_server_access(prisma_client, user_api_key_dict, server_id)
         # Filter to only known per-user var names declared by the admin —
         # never persist arbitrary keys the user invents.
         _, user_specs = parse_admin_env_vars(getattr(server, "env_vars", None))
@@ -2316,13 +2316,13 @@ if MCP_AVAILABLE:
                 status_code=status.HTTP_400_BAD_REQUEST,
                 detail={"error": "User ID not found in token"},
             )
+        await _authorize_mcp_server_access(prisma_client, user_api_key_dict, server_id)
         server = await get_mcp_server(prisma_client, server_id)
         if server is None:
             raise HTTPException(
                 status_code=status.HTTP_404_NOT_FOUND,
                 detail={"error": f"MCP Server {server_id} not found"},
             )
-        await _authorize_mcp_server_access(prisma_client, user_api_key_dict, server_id)
         await delete_user_env_vars(prisma_client, user_id, server_id)
         return _compute_user_env_var_status(server=server, stored_values={})
 

tests/test_litellm/proxy/management_endpoints/test_mcp_management_endpoints.py

@@ -4141,3 +4141,31 @@ class TestMCPUserEnvVarsAccessControl:
             )
         assert result.server_id == "srv-1"
         access_list_mock.assert_not_awaited()
+
+    @pytest.mark.asyncio
+    async def test_non_admin_gets_403_not_404_for_inaccessible_server(self):
+        """Authorization must run before the existence check so a non-admin
+        cannot distinguish "server does not exist" (404) from "server exists but
+        you lack access" (403) and enumerate server IDs."""
+        get_mcp_server_mock = AsyncMock(return_value=None)
+        with (
+            patch.object(
+                mgmt_endpoints, "get_prisma_client_or_throw", return_value=MagicMock()
+            ),
+            patch.object(mgmt_endpoints, "get_mcp_server", get_mcp_server_mock),
+            patch.object(
+                mgmt_endpoints,
+                "get_all_mcp_servers_for_user",
+                AsyncMock(return_value=[]),
+            ),
+        ):
+            with pytest.raises(HTTPException) as exc:
+                await mgmt_endpoints.get_mcp_user_env_vars(
+                    server_id="srv-1",
+                    user_api_key_dict=generate_mock_user_api_key_auth(
+                        user_id="alice",
+                        user_role=LitellmUserRoles.INTERNAL_USER,
+                    ),
+                )
+        assert exc.value.status_code == 403
+        get_mcp_server_mock.assert_not_awaited()

ui/litellm-dashboard/src/components/networking.tsx

@@ -10119,7 +10119,7 @@ export const listMCPUserCredentials = async (
 export const getMCPUserEnvVars = async (
   accessToken: string,
   serverId: string,
-): Promise<MCPUserEnvVarsStatus | null> => {
+): Promise<MCPUserEnvVarsStatus> => {
   const url = proxyBaseUrl
     ? `${proxyBaseUrl}/v1/mcp/server/${serverId}/user-env-vars`
     : `/v1/mcp/server/${serverId}/user-env-vars`;
@@ -10127,7 +10127,15 @@ export const getMCPUserEnvVars = async (
     method: "GET",
     headers: { [globalLitellmHeaderName]: `Bearer ${accessToken}` },
   });
-  if (!response.ok) return null;
+  if (!response.ok) {
+    const err = await response.json().catch(() => ({}));
+    const detail = (err as { detail?: unknown })?.detail;
+    const message =
+      typeof detail === "string"
+        ? detail
+        : (detail as { error?: string })?.error || "Failed to load env vars";
+    throw new Error(message);
+  }
   return response.json();
 };