This is how the non-error/exc response looks like now:

[{"type": "text", "text": "The weather in Seattle, WA is sunny.", "annotations": null}]

And an error response/exception (the text value is coming from the mcp server tool):

[{"type": "text", "text": "Error occured while fetching weather", "annotations": null}]

Codecov Report

Attention: Patch coverage is 89.65517% with 3 lines in your changes missing coverage. Please review.

Project coverage is 79.29%. Comparing base (9118f9b) to head (dbc0adb).
Report is 2 commits behind head on main.

@@            Coverage Diff             @@
##             main    #6482      +/-   ##
==========================================
+ Coverage   79.23%   79.29%   +0.06%     
==========================================
  Files         225      225              
  Lines       16666    16679      +13     
==========================================
+ Hits        13205    13226      +21     
+ Misses       3461     3453       -8     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

Hi @peterj ,

Can you add a minimum repro example in the PR description, thanks.

Edit: able to repro

# server.py 
import logging
from mcp.server.fastmcp import FastMCP
 
# Create an MCP server
mcp = FastMCP("Demo")
 

# create an error agent that raises an error
@mcp.tool() 
async def error_agent(task: str) -> str:
    """An agent that raises an error"""
    raise Exception("This is an error agent") 

if __name__ == "__main__":
    mcp.run()

from autogen_ext.models.openai import OpenAIChatCompletionClient
from autogen_ext.tools.mcp import StdioServerParams, mcp_server_tools
from autogen_agentchat.agents import AssistantAgent
from autogen_core import CancellationToken
from autogen_agentchat.ui import Console

async def main() -> None:
    # Setup server params for local filesystem access
    fetch_mcp_server = StdioServerParams(command="/pathtopython/python", args=["pathtoserver/server.py"])
    tools = await mcp_server_tools(fetch_mcp_server)

    # Create an agent that can use the fetch tool.
    model_client = OpenAIChatCompletionClient(model="gpt-4o")
    agent = AssistantAgent(name="fetcher", model_client=model_client, tools=tools, reflect_on_tool_use=True)   

    await Console(agent.run_stream(task="raise an error withthe errror agent", cancellation_token=CancellationToken()))
    
await main()

will fix these

one of these days I'll be able to fix all lining errors :)

Haha,
I can take over from here (help with final test/lint fixes, merge conflicts).
Do you mind giving me push access to your repo (https://github.com/peterj/autogen)?
I seem to be getting auth issues when I try to update this PR (possibly related to git lfs).

This is mostly ready to get merged, thanks again!

Sorry I mention this here, I encountered a problem when using serialization that would cause Chinese programming unicode encoding, which would make the print look weird (although the result is correct)
I can fix it by making local code modifications：

return json.dumps([serialize_item(item) for item in value]， ensure_ascil=False)

However, it may not be very convenient for service deployment and packaging, as repackaging would require obtaining new dependencies again.

In the context of building a multi-agent system, multiple rounds of conversation history may introduce Unicode-encoded text, which could potentially affect model inference — though the exact impact is hard to quantify.

Is this something that will be addressed in future versions?