Thanks to visit codestin.com
Credit goes to github.com

Skip to content

[HELP] Error listing prompts for connector streamable HTTP and LangGraph blocking #626

@thanhtung2693

Description

@thanhtung2693

Hello, I've just started using mcp-use with langgraph/langchain but got 2 issues.
Don't know if I missed anything from the document.

Error listing prompts for connector streamable HTTP

First I create new server using this template: https://mcp-use.com/docs/typescript/server/templates#%F0%9F%8E%A8-mcp-ui-template

After server started, I use with langchain adapter:

config = {
    "mcpServers": {
        "uiresource-mcp-server": {
            "url": "http://localhost:3000/mcp"
        },
    }
}

try:
    client = MCPClient(config=config)

    # Creates the adapter for LangChain's format
    adapter = LangChainAdapter()

    # Convert tools from active connectors to the LangChain's format
    await adapter.create_all(client)

    # List concatenation (if you loaded all tools)
    langchain_tools = adapter.tools + adapter.resources + adapter.prompts

    # Create chat model
    model = ChatOllama(
        model="qwen3-vl:2b-instruct-q4_K_M",
        temperature=0.2
    )
    # Create the LangChain agent
    agent = create_agent(
        model=model,
        tools=langchain_tools,
        system_prompt="You are a helpful assistant",
    )

    # Run the agent
    result = await agent.ainvoke(
        {
            "messages": [
                {
                    "role": "user",
                    "content": "use welcome-card",
                }
            ]
        }
    )
except Exception as e:
    print(f"Error: {e}")
    raise e

and here is the console output:

2025-12-04 23:57:56,587 - mcp_use - INFO - No active sessions found, creating new ones...
2025-12-04 23:57:56,595 - mcp_use - INFO - Using default callback port 8080
2025-12-04 23:57:56,623 - mcp_use - ERROR - Failed to discover OAuth/OIDC metadata for http://localhost:3000/mcp
2025-12-04 23:57:57,065 - mcp_use - ERROR - Error listing prompts for connector streamable HTTP:http://localhost:3000/mcp: Method not found
Welcome to MCP-UI! Your server is running and ready to serve interactive widgets!

Here are some details:
- Widget Types: 3
- Possibilities: ∞
- Fast & Simple: ⚡

Server: **uiresource-mcp-server v1.0.0**  
Port: **3000**

Enjoy the experience!

Despite the error, the LLM still can answer and return the HTML for welcome-card

LangGraph Blocking call

Seem like MCP Client causing this error when authen, here is my code

async def graph():
    config = {
        "mcpServers": {
            "uiresource-mcp-server": {
                "url": "http://localhost:3000/mcp"
            }
        }
    }
    
    client = MCPClient(config=config)

    # Creates the adapter for LangChain's format
    adapter = LangChainAdapter()

    # Convert tools from active connectors to the LangChain's format
    await adapter.create_tools(client)

    # List concatenation (if you loaded all tools)
    langchain_tools = adapter.tools
    
    print(langchain_tools)

    # Create chat model
    model = ChatOllama(
        model="qwen3-vl:2b-instruct-q4_K_M",
        temperature=0.2
    )
    
    # Create agent with MCP tools
    agent = create_agent(
        model=model,
        tools=langchain_tools
    )
    
    return agent

And got this error:

self.token_storage = token_storage or FileTokenStorage()

blockbuster.blockbuster.BlockingError: Blocking call to os.mkdir

Heads up! LangGraph dev identified a synchronous blocking call in your code. When running in an ASGI web server, blocking calls can degrade performance for everyone since they tie up the event loop.

Here are your options to fix this:

1. Best approach: Convert any blocking code to use async/await patterns
   For example, use 'await aiohttp.get()' instead of 'requests.get()'

2. Quick fix: Move blocking operations to a separate thread
   Example: 'await asyncio.to_thread(your_blocking_function)'

3. Override (if you can't change the code):
   - For development: Run 'langgraph dev --allow-blocking'
   - For deployment: Set 'BG_JOB_ISOLATED_LOOPS=true' environment variable

These blocking operations can prevent health checks and slow down other runs 

Can anyone help, thank you!!!

Metadata

Metadata

Labels

PythonagentbugSomething isn't workingclientdocumentationImprovements or additions to documentationreadyWhen a PR is ready to be merged and waiting for author to mergeserver

Type

No type

Projects

No projects

Milestone

No milestone

Relationships

None yet

Development

No branches or pull requests

Issue actions