Repository navigation
azure-ai-agents: MCP usage errors with Azure AI Foundry #42428
Description
Activity
Also, when the tool is successfully invoked, accomplished after changing the prompt to be more enforcing, I am still seeing a failure result:
"detail": "An unexpected error occurred: Failed to process query: Agent run failed: server_error - Sorry, something went wrong.",This is the only context I receive in the logs which isn't particularly insightful. I can remote debug into your SDK, although more insightful logging would be ideal.
Thanks for the feedback! We are routing this to the appropriate team for follow-up. cc Darren Cohen (@dargilco) Glenn Harper (@glharper) Howie Leung (@howieleung) Jarno Hakulinen (@jhakulin) Nikolay Rovinskiy (@nick863) Travis Angevine (@trangevi).
reesmonty (@jacobreesmontgomery) We are aware of multiple issues involving MCP tool usage in Agents and are triaging them now.
I’m facing the same issue. Is there any update or workaround available?
32 remaining items
ith
npx:npx @modelcontextprotocol/inspecWhat are the exact steps you follow to get this working? I could not figure it out.
@jacobreesmontgomery Does submitting an approval not work, or does disabling approval - setting
require_approvaltonever- not work? Can you share your SDK example?Setting
require_approvaltoneverdoes not work, seemingly.I shared the example above. Not sure what you mean.
reesmonty (@jacobreesmontgomery) can you please share your Python code snippet that doesn't work?
Reacted by Glenn HarperAlex Pryiomka (@apryiomka) , I'd be happy to hop on a call to briefly show you, but I'm not comfortable disclosing that here since this is FedEx-internal code.
Reacted by Alex Pryiomka@apryiomka , I'd be happy to hop on a call to briefly show you, but I'm not comfortable disclosing that here since this is FedEx-internal code.
I can take a look with you, reesmonty (@jacobreesmontgomery). My email is <my_gh_username>(at)microsoft(dot)com
Reacted by reesmontyI have the same issue. I can see that although listcall works, call_tool fails since it terminates the session before the tool_Call. I ran wireshark to analyze the packets and I see DELETE being called before calling the tool. I have tested the MCP server using FastMCP client and MS Copilot Studio- and in both situations, my MCP server works fine. Azure AI Agent SDK however fails.
Python:
Name: azure-ai-agents
Version: 1.2.0b3Below is my code that I am using.
`import os, time
#import gradio as gr
from dotenv import load_dotenv
import asyncioAzure AI Agent SDK
from azure.ai.projects.aio import AIProjectClient
from azure.identity.aio import DefaultAzureCredential
from azure.ai.agents.models import (
ListSortOrder,
McpTool,
RequiredMcpToolCall,
RunStepActivityDetails,
SubmitToolApprovalAction,
ToolApproval,
)Load environment variables
load_dotenv()
Azure Agent + OpenAI configuration
endpoint = os.getenv("AZ_PROJ_ENDPNT") # Azure project endpoint (Agent Service)
api_key = os.getenv("AZ_OPENAPI") # Your Azure OpenAI / Agent key
deployment = os.getenv("AZ_MODEL") # Model deployment name
mcp_server_url = os.getenv("MCP_SERVER_URL")
mcp_server_label = os.getenv("MCP_SERVER_LABEL")
agent_id = os.getenv("AZ_AGENT_ID")async def main() -> None:
project_client = AIProjectClient( endpoint=endpoint, credential=DefaultAzureCredential(), ) # Initialize agent MCP tool mcp_tool = McpTool( server_label=mcp_server_label, server_url=mcp_server_url, #allowed_tools=[], # Optional: specify allowed tools ) mcp_tool.set_approval_mode("never") # Uncomment to disable approval requirement async with project_client: agents_client = project_client.agents agent = await agents_client.get_agent(agent_id=agent_id) # print(f"Created agent, ID: {agent.id}") # print(f"MCP Server: {mcp_tool.server_label} at {mcp_tool.server_url}") #Create thread for communication thread = await agents_client.threads.create() print(f"Created thread, ID: {thread.id}") # Create message to thread message = await agents_client.messages.create( thread_id=thread.id, role="user", content="Please list all the devices under company Example", ) print(f"Created message, ID: {message.id}") #Create and process agent run in thread with MCP tools #mcp_tool.update_headers("SuperSecret", "123456") run = await agents_client.runs.create(thread_id=thread.id, agent_id=agent.id, tool_resources=mcp_tool.resources) print(f"Created run, ID: {run.id}") while run.status in ["queued", "in_progress", "requires_action"]: await asyncio.sleep(1) run = await agents_client.runs.get(thread_id=thread.id, run_id=run.id) if run.status == "requires_action" and isinstance(run.required_action, SubmitToolApprovalAction): tool_calls = run.required_action.submit_tool_approval.tool_calls if not tool_calls: print("No tool calls provided - cancelling run") await agents_client.runs.cancel(thread_id=thread.id, run_id=run.id) break tool_approvals = [] for tool_call in tool_calls: if isinstance(tool_call, RequiredMcpToolCall): try: print(f"Approving tool call: {tool_call}") tool_approvals.append( ToolApproval( tool_call_id=tool_call.id, approve=True, headers=mcp_tool.headers, ) ) except Exception as e: print(f"Error approving tool_call {tool_call.id}: {e}") print(f"tool_approvals: {tool_approvals}") if tool_approvals: await agents_client.runs.submit_tool_outputs( thread_id=thread.id, run_id=run.id, tool_approvals=tool_approvals ) print(f"Current run status: {run.status}") print(f"Run completed with status: {run.status}") if run.status == "failed": print(f"Run failed: {run.last_error}") # Display run steps and tool calls run_steps = agents_client.run_steps.list(thread_id=thread.id, run_id=run.id) # Loop through each step async for step in run_steps: print(f"Step {step['id']} status: {step['status']}") # Check if there are tool calls in the step details step_details = step.get("step_details", {}) tool_calls = step_details.get("tool_calls", []) if tool_calls: print(" MCP Tool calls:") for call in tool_calls: print(f" Tool Call ID: {call.get('id')}") print(f" Type: {call.get('type')}") if isinstance(step_details, RunStepActivityDetails): for activity in step_details.activities: for function_name, function_definition in activity.tools.items(): print( f' The function {function_name} with description "{function_definition.description}" will be called.:' ) if len(function_definition.parameters) > 0: print(" Function parameters:") for argument, func_argument in function_definition.parameters.properties.items(): print(f" {argument}") print(f" Type: {func_argument.type}") print(f" Description: {func_argument.description}") else: print("This function has no parameters") print() # add an extra newline between steps # Fetch and log all messages messages = agents_client.messages.list(thread_id=thread.id, order=ListSortOrder.ASCENDING) print("\nConversation:") print("-" * 50) async for msg in messages: if msg.text_messages: last_text = msg.text_messages[-1] print(f"{msg.role.upper()}: {last_text.text.value}") print("-" * 50) # Example of dynamic tool management # print(f"\nDemonstrating dynamic tool management:") # print(f"Current allowed tools: {mcp_tool.allowed_tools}") # Clean-up and delete the agent once the run is finished. # NOTE: Comment out this line if you plan to reuse the agent later. # await agents_client.delete_agent(agent.id) # print("Deleted agent")if name == "main":
asyncio.run(main())`Screenshots:
- DELETE:
- Tool Call (after DELETE)
- Failed 404:
Notes:
My MCP server is accessible online but access is restricted only from MS published IP list. Therefore not accessible to general public to test with.
I have tested the tool calling with RestAPI and its the same result.@apryiomka , I'd be happy to hop on a call to briefly show you, but I'm not comfortable disclosing that here since this is FedEx-internal code.
reesmonty (@jacobreesmontgomery) could you please email to oai-assistants@microsoft.com, we will tirage on our end. We can continue the conversation over the email.
reesmonty (@jacobreesmontgomery) please also note that the tool resource / approval has to match on the
server_labelproperty.@apryiomka , I'd be happy to hop on a call to briefly show you, but I'm not comfortable disclosing that here since this is FedEx-internal code.
@jacobreesmontgomery could you please email to oai-assistants@microsoft.com, we will tirage on our end. We can continue the conversation over the email.
Alex Pryiomka (@apryiomka) sorry for my case,
but would this addressed as part of a bug or would you want me to open a new case. I believe its the same issue. I have posted my findings aboveHi reesmonty (@jacobreesmontgomery). Thank you for opening this issue and giving us the opportunity to assist. We believe that this has been addressed. If you feel that further discussion is needed, please add a comment with the text "/unresolve" to remove the "issue-addressed" label and continue the conversation.
Hi reesmonty (@jacobreesmontgomery), since you haven’t asked that we
/unresolvethe issue, we’ll close this out. If you believe further discussion is needed, please add a comment/unresolveto reopen the issue.- locked and limited conversation to collaborators
on Dec 9, 2025
azure-ai-projects: 1.0.0b10azure-ai-agents: 1.2.0b1Describe the bug
I've been having issues using the MCP integrations with Azure AI Foundry agents. Specifically, there are two things:
AssertionErroris frequently occurring. It's happened for both GitMCP and Tavily's MCP servers. I have yet to try more MCP servers, but I suspect I'll continue to see these errors coming up. See the error below.Error to problem 1
The result with the error:
{ "data": { "response": "I encountered an issue while trying to get the current weather information for Seattle. You can check a reliable weather website or app for real-time updates. Alternatively, if you need general information about Seattle's climate or typical weather patterns, feel free to ask!" }, "meta": { "run_steps": [ { "id": "step_hgixrj0R0tORvoMZpTkf14cN", "object": "thread.run.step", "created_at": 1754658900, "run_id": "run_cT1NcPI13R3b88JJA9oDq179", "assistant_id": "asst_GyjPbDs3XfMbSrv0VehZmcFv", "thread_id": "thread_jlp4vdxxi3HlI7F5E67A6jEC", "type": "message_creation", "status": "completed", "cancelled_at": null, "completed_at": 1754658901, "expires_at": null, "failed_at": null, "last_error": null, "step_details": { "type": "message_creation", "message_creation": { "message_id": "msg_q8TpSGSexr47v7Pb0nImYnbc" } }, "usage": { "prompt_tokens": 152, "completion_tokens": 52, "total_tokens": 204, "prompt_token_details": { "cached_tokens": 0 } } }, { "id": "step_ovu486RNvTcTKZ0zIBUmEB72", "object": "thread.run.step", "created_at": 1754658900, "run_id": "run_cT1NcPI13R3b88JJA9oDq179", "assistant_id": "asst_GyjPbDs3XfMbSrv0VehZmcFv", "thread_id": "thread_jlp4vdxxi3HlI7F5E67A6jEC", "type": "tool_calls", "status": "completed", "cancelled_at": null, "completed_at": 1754658900, "expires_at": null, "failed_at": null, "last_error": null, "step_details": { "type": "tool_calls", "tool_calls": [ { "id": "call_UWl9nlG2w8akjCWKLMpgdeKi", "type": "mcp", "arguments": "{\"location\":\"Seattle, WA\"}", "name": "weather", "server_label": "tavily", "output": "content_type='system_error' name='AssertionError' text=\"Encountered exception: <class 'AssertionError'>.\"" } ] }, "usage": { "prompt_tokens": 106, "completion_tokens": 19, "total_tokens": 125, "prompt_token_details": { "cached_tokens": 0 } } } ] } }Error to problem 2
{ "data": { "response": "I currently don't have real-time capabilities to check the latest weather. However, you can easily find the current weather in Seattle by checking a weather website or app like Weather.com, the Weather Channel, or a local news website. Alternatively, you can use a digital assistant on your phone or smart device to get the latest weather update." }, "meta": { "run_steps": [ { "id": "step_47k5HIxdJe4D1T02IvGfzqZ5", "object": "thread.run.step", "created_at": 1754659678, "run_id": "run_HF0DgO0m9n8PhVZyYqMKz7JL", "assistant_id": "asst_xOCNz0JcZ2yMbzbDDZ96FIee", "thread_id": "thread_uKI5zjsP5xAIhXfJ9rEO0x43", "type": "message_creation", "status": "completed", "cancelled_at": null, "completed_at": 1754659679, "expires_at": null, "failed_at": null, "last_error": null, "step_details": { "type": "message_creation", "message_creation": { "message_id": "msg_ThsZroUC9qje77hFJuPTuSR8" } }, "usage": { "prompt_tokens": 106, "completion_tokens": 68, "total_tokens": 174, "prompt_token_details": { "cached_tokens": 0 } } } ] } }To Reproduce
Steps to reproduce the behavior:
server_label: "tavily"server_url: "https://mcp.tavily.com/mcp/?tavilyApiKey=<MY_TAVILY_API_KEY>",allowed_tools: ["tavily-search"]Expected behavior
I would expect two things:
tavily-searchtool is invoked.Screenshots
N/A
Additional context
gpt-4owith a provided temperature of0.7.You are a helpful assistant. When appropriate, use the available tools to help answer the user's query..