Fix MCP server support using MCPStreamableHTTPTool

README.md

@@ -73,6 +73,43 @@ Enter a message like `what are the laws?`
 
 The application will start an interactive conversation loop where you can ask questions. Type 'exit' or 'quit' to end the session.
 
+### Using MCP Servers
+
+The application supports Model Context Protocol (MCP) servers for extending agent capabilities with remote tools. To use an MCP server:
+
+1. Uncomment one of the MCP tool examples in `main.py`:
+
+   **Example 1 - Microsoft Learn MCP server:**
+   ```python
+   mcp_tool = MCPStreamableHTTPTool(
+       name="learn-mcp",
+       url="https://learn.microsoft.com/mcp",
+       description="Microsoft Learn MCP server"
+   )
+   ```
+
+   **Example 2 - GitHub MCP server with OAuth:**
+   ```python
+   mcp_tool = MCPStreamableHTTPTool(
+       name="github-mcp",
+       url="https://api.github.com/mcp",
+       headers={"Authorization": f"Bearer {os.getenv('GITHUB_TOKEN')}"},
+       description="GitHub MCP server"
+   )
+   ```
+
+2. Uncomment the async context manager line:
+   ```python
+   # async with mcp_tool:
+   ```
+
+3. Uncomment the tools parameter in the ChatAgent:
+   ```python
+   # , tools=mcp_tool
+   ```
+
+For more information about MCP tools, see the [Agent Framework MCP documentation](https://learn.microsoft.com/en-us/agent-framework/user-guide/model-context-protocol/using-mcp-tools?pivots=programming-language-python).
+
 ## Authentication
 
 The application uses `DefaultAzureCredential` for authentication. Make sure you're logged in to Azure CLI:

main.py

@@ -1,6 +1,6 @@
 import asyncio
 import os
-from agent_framework import ChatAgent
+from agent_framework import ChatAgent, MCPStreamableHTTPTool
 from agent_framework.azure import AzureOpenAIChatClient
 from azure.identity import DefaultAzureCredential, get_bearer_token_provider
 
@@ -35,23 +35,38 @@
     ad_token_provider=token_provider
 )
 
-# Create the agent
-agent = ChatAgent(
-    chat_client=chat_client,
-    instructions=instructions
-)
-
-# Optional: Add MCP tool from remote URL
-# agent.add_mcp_server("https://example.com/mcp-server")
+# Optional: Create MCP tool from remote URL
+# Example 1 - Microsoft Learn MCP server:
+# mcp_tool = MCPStreamableHTTPTool(
+#     name="learn-mcp",
+#     url="https://learn.microsoft.com/mcp",
+#     description="Microsoft Learn MCP server"
+# )
 
-# Optional: Add MCP tool with authentication
-# import httpx
-# headers = {"Authorization": "Bearer YOUR_TOKEN_HERE"}
-# agent.add_mcp_server("https://example.com/mcp-server", headers=headers)
+# Example 2 - GitHub MCP server with OAuth:
+# mcp_tool = MCPStreamableHTTPTool(
+#     name="github-mcp",
+#     url="https://api.github.com/mcp",
+#     headers={"Authorization": f"Bearer {os.getenv('GITHUB_TOKEN')}"},
+#     description="GitHub MCP server"
+# )
 
 # Stay in a loop for continuous conversation
 async def main():
     """Run interactive conversation loop with the agent."""
+
+    # Optional: Use MCP tool in async context manager
+    # Uncomment to use MCP tools with the agent:
+    # async with mcp_tool:
+
+    # Create the agent
+    agent = ChatAgent(
+        chat_client=chat_client,
+        instructions=instructions
+        # Uncomment to add MCP tools:
+        # , tools=mcp_tool
+    )
+
     while True:
         user_message = input("Enter your message: ")