diff --git a/integration-tests/interactive/protocol-tags-interactive.test.ts b/integration-tests/interactive/protocol-tags-interactive.test.ts
index 73c2ebc4be..a45a7a238a 100644
--- a/integration-tests/interactive/protocol-tags-interactive.test.ts
+++ b/integration-tests/interactive/protocol-tags-interactive.test.ts
@@ -12,52 +12,85 @@ import {
 } from '../fake-openai-server.js';
 import { TestRig, type } from '../test-helper.js';
 
+const SANDBOX_MODE = process.env['QWEN_SANDBOX']?.toLowerCase().trim();
+const IS_CONTAINER_SANDBOX =
+  SANDBOX_MODE === 'docker' || SANDBOX_MODE === 'podman';
+
 describe('Interactive protocol tag retry guard', () => {
   let fakeServer: FakeOpenAIServer | undefined;
   let rig: TestRig;
+  let savedNoProxy: string | undefined;
+  let savedNoProxyLower: string | undefined;
 
   beforeEach(() => {
     rig = new TestRig();
+    if (IS_CONTAINER_SANDBOX) {
+      savedNoProxy = process.env['NO_PROXY'];
+      savedNoProxyLower = process.env['no_proxy'];
+      const noProxy = '127.0.0.1,localhost,host.docker.internal';
+      process.env['NO_PROXY'] = noProxy;
+      process.env['no_proxy'] = noProxy;
+    }
   });
 
   afterEach(async () => {
     await fakeServer?.close();
     fakeServer = undefined;
+    if (IS_CONTAINER_SANDBOX) {
+      if (savedNoProxy !== undefined) {
+        process.env['NO_PROXY'] = savedNoProxy;
+      } else {
+        delete process.env['NO_PROXY'];
+      }
+      if (savedNoProxyLower !== undefined) {
+        process.env['no_proxy'] = savedNoProxyLower;
+      } else {
+        delete process.env['no_proxy'];
+      }
+    }
     await rig.cleanup();
   });
 
   it.skipIf(process.platform === 'win32')(
     'retries protocol leaks across SSE disconnect and completed streams',
     async () => {
-      fakeServer = await startFakeOpenAIServer(({ requestIndex }) => {
-        if (requestIndex === 0) {
-          return {
-            contentChunks: [
-              '<analysis>hidden before disconnect',
-              '</analysis><summary>WRONG_FIRST_ATTEMPT</summary>',
-            ],
-            disconnectAfterContentChunks: 1,
-          };
-        }
+      fakeServer = await startFakeOpenAIServer(
+        ({ requestIndex }) => {
+          if (requestIndex === 0) {
+            return {
+              contentChunks: [
+                '<analysis>hidden before disconnect',
+                '</analysis><summary>WRONG_FIRST_ATTEMPT</summary>',
+              ],
+              disconnectAfterContentChunks: 1,
+            };
+          }
 
-        if (requestIndex === 1) {
-          return {
-            contentChunks: [
-              '<analysis>hidden completed attempt</analysis>',
-              '<summary>WRONG_COMPLETED_SUMMARY</summary>',
-            ],
-          };
-        }
+          if (requestIndex === 1) {
+            return {
+              contentChunks: [
+                '<analysis>hidden completed attempt</analysis>',
+                '<summary>WRONG_COMPLETED_SUMMARY</summary>',
+              ],
+            };
+          }
 
-        return {
-          contentChunks: ['VISIBLE_TMUX_RETRY_RESPONSE_DONE'],
-          usage: {
-            prompt_tokens: 20,
-            completion_tokens: 8,
-            total_tokens: 28,
-          },
-        };
-      });
+          return {
+            contentChunks: ['VISIBLE_TMUX_RETRY_RESPONSE_DONE'],
+            usage: {
+              prompt_tokens: 20,
+              completion_tokens: 8,
+              total_tokens: 28,
+            },
+          };
+        },
+        IS_CONTAINER_SANDBOX
+          ? {
+              listenHost: '0.0.0.0',
+              baseUrlHost: 'host.docker.internal',
+            }
+          : undefined,
+      );
 
       await rig.setup('interactive-protocol-tag-filtering-http-retry', {
         settings: {