Updated the test cases to support the headers for now

src/llama_stack/providers/remote/tool_runtime/model_context_protocol/model_context_protocol.py

@@ -56,9 +56,7 @@ class ModelContextProtocolToolRuntimeImpl(ToolGroupsProtocolPrivate, ToolRuntime
         final_authorization = authorization or provider_auth
 
         return await list_mcp_tools(
-            endpoint=mcp_endpoint.uri,
-            headers=provider_headers,
-            authorization=final_authorization
+            endpoint=mcp_endpoint.uri, headers=provider_headers, authorization=final_authorization
         )
 
     async def invoke_tool(
@@ -109,7 +107,7 @@ class ModelContextProtocolToolRuntimeImpl(ToolGroupsProtocolPrivate, ToolRuntime
         authorization = None
 
         provider_data = self.get_request_provider_data()
-        if provider_data and hasattr(provider_data, 'mcp_headers') and provider_data.mcp_headers:
+        if provider_data and hasattr(provider_data, "mcp_headers") and provider_data.mcp_headers:
             for uri, values in provider_data.mcp_headers.items():
                 if canonicalize_uri(uri) != canonicalize_uri(mcp_endpoint_uri):
                     continue

tests/integration/inference/test_tools_with_schemas.py

@@ -9,6 +9,8 @@ Integration tests for inference/chat completion with JSON Schema-based tools.
 Tests that tools pass through correctly to various LLM providers.
 """
 
+import json
+
 import pytest
 
 from llama_stack.core.library_client import LlamaStackAsLibraryClient
@@ -191,9 +193,22 @@ class TestMCPToolsInChatCompletion:
             mcp_endpoint=dict(uri=uri),
         )
 
+        # Use old header-based approach for Phase 1 (backward compatibility)
+        provider_data = {
+            "mcp_headers": {
+                uri: {
+                    "Authorization": f"Bearer {AUTH_TOKEN}",
+                },
+            },
+        }
+        auth_headers = {
+            "X-LlamaStack-Provider-Data": json.dumps(provider_data),
+        }
+
         # Get the tools from MCP
         tools_response = llama_stack_client.tool_runtime.list_tools(
             tool_group_id=test_toolgroup_id,
+            extra_headers=auth_headers,
         )
 
         # Convert to OpenAI format for inference