diff --git a/src/auto-reply/reply/conversation-label-generator.test.ts b/src/auto-reply/reply/conversation-label-generator.test.ts
index 819f27f44b5..4b152391efe 100644
--- a/src/auto-reply/reply/conversation-label-generator.test.ts
+++ b/src/auto-reply/reply/conversation-label-generator.test.ts
@@ -135,6 +135,40 @@ describe("generateConversationLabel", () => {
     expect(call[2].signal).toBeInstanceOf(AbortSignal);
   });
 
+  it("applies prepared runtime auth to the completion model", async () => {
+    resolveDefaultModelForAgent.mockReturnValue({
+      provider: "microsoft-foundry",
+      model: "claude-fable-5",
+    });
+    resolveModelAsync.mockResolvedValue({
+      model: {
+        provider: "microsoft-foundry",
+        api: "anthropic-messages",
+        baseUrl: "https://example.services.ai.azure.com/anthropic",
+        headers: { "x-api-key": "stale-key" },
+      },
+      authStorage: {},
+      modelRegistry: {},
+    });
+    getRuntimeAuthForModel.mockResolvedValue({
+      apiKey: "entra-token",
+      mode: "api-key",
+      request: { auth: { mode: "authorization-bearer", token: "entra-token" } },
+    });
+    requireApiKey.mockReturnValue("entra-token");
+
+    await generateConversationLabel({
+      userMessage: "Need help with invoices",
+      prompt: "Generate a label",
+      cfg: {},
+    });
+
+    expect(requireFirstMockCall(completeSimple, "simple completion")[0]).toMatchObject({
+      provider: "microsoft-foundry",
+      headers: { Authorization: "Bearer entra-token" },
+    });
+  });
+
   it("omits temperature for Codex Responses simple completions", async () => {
     resolveDefaultModelForAgent.mockReturnValue({ provider: "openai", model: "gpt-5.5" });
     resolveModelAsync.mockResolvedValue({
diff --git a/src/auto-reply/reply/conversation-label-generator.ts b/src/auto-reply/reply/conversation-label-generator.ts
index 9516da431b6..39ef493886e 100644
--- a/src/auto-reply/reply/conversation-label-generator.ts
+++ b/src/auto-reply/reply/conversation-label-generator.ts
@@ -2,6 +2,7 @@
 import { resolveModelAsync } from "../../agents/embedded-agent-runner/model.js";
 import { requireApiKey } from "../../agents/model-auth.js";
 import { resolveDefaultModelForAgent } from "../../agents/model-selection.js";
+import { applyPreparedRuntimeAuthToModel } from "../../agents/provider-request-config.js";
 import { prepareModelForSimpleCompletion } from "../../agents/simple-completion-transport.js";
 import type { OpenClawConfig } from "../../config/types.openclaw.js";
 import { logVerbose } from "../../globals.js";
@@ -61,21 +62,20 @@ export async function generateConversationLabel(
   }
   const completionModel = prepareModelForSimpleCompletion({ model: resolved.model, cfg });
 
-  const apiKey = requireApiKey(
-    await getRuntimeAuthForModel({
-      model: completionModel,
-      cfg,
-      workspaceDir: agentDir,
-    }),
-    modelRef.provider,
-  );
+  const runtimeAuth = await getRuntimeAuthForModel({
+    model: completionModel,
+    cfg,
+    workspaceDir: agentDir,
+  });
+  const apiKey = requireApiKey(runtimeAuth, modelRef.provider);
+  const runtimeModel = applyPreparedRuntimeAuthToModel(completionModel, runtimeAuth);
 
   const controller = new AbortController();
   const timeout = setTimeout(() => controller.abort(), TIMEOUT_MS);
   try {
     // Label generation should never block normal reply handling for long.
     const result = await completeSimple(
-      completionModel,
+      runtimeModel,
       {
         systemPrompt: prompt,
         messages: [
@@ -89,7 +89,7 @@ export async function generateConversationLabel(
       {
         apiKey,
         maxTokens: 100,
-        ...(isCodexSimpleCompletionModel(completionModel) ? {} : { temperature: 0.3 }),
+        ...(isCodexSimpleCompletionModel(runtimeModel) ? {} : { temperature: 0.3 }),
         signal: controller.signal,
       },
     );