[Bug]: Cannot invoke Virtual Server tools using LangChain

[gfollower]

🐞 Bug Summary

I am unable to call tools in my Virtual Server based on the official langchain_mcp_adapters documentation:
https://docs.langchain.com/oss/python/langchain/mcp#model-context-protocol-mcp

---

🧩 Affected Component

Select the area of the project impacted:

• mcpgateway - API
• mcpgateway - UI (admin panel)
• mcpgateway.wrapper - stdio wrapper
• Federation or Transports
• CLI, Makefiles, or shell scripts
• Container setup (Docker/Podman/Compose)
• Other (explain below)

---

🔁 Steps to Reproduce

Env:
fastmcp 2.11.3
langchain-mcp-adapters 0.1.14
langgraph 1.0.2

The issue can be reproduced by following the steps below:

• Create a basic MCP server using FastMCP

from mcp.server.fastmcp import FastMCP

mcp = FastMCP("Weather", port=8010, host="0.0.0.0")

@mcp.tool()
async def get_weather(location: str) -> dict:
    """Get weather for location."""
    return {"result": "It's always sunny in New York"}

if __name__ == "__main__":
    mcp.run(transport="streamable-http")

• Register this mcp server to gateway, and create a pulic Virtual Server with this tool, then we get the server url:

http://{docker_hostname}:4444/servers/4fec000a5ec2456aae2c71e0f2920311/mcp

-Use my Virtual Server url in LangChain

from langchain_mcp_adapters.client import MultiServerMCPClient
from langchain.agents import create_agent

async def main():
    client = MultiServerMCPClient(
        {
            "gateway": {
                "transport": "streamable_http",
                "url": f"http://{docker_hostname}:4444/servers/4fec000a5ec2456aae2c71e0f2920311/mcp",
                "headers": {
                    "Authorization": f"Bearer {jwt_token}"
                }
            }
        }
    )

    tools = await client.get_tools()
    agent = create_agent(
        llm,
        tools
    )
    response = await agent.ainvoke({"messages": [{"role": "user", "content": "what is the weather in nyc?"}]})
    print(response)

asyncio.run(main())

After running, I got the following response...The tool_calls were not actually executed, and no ToolMessage was generated:

{'messages': [HumanMessage(content='what is the weather in nyc?', additional_kwargs={}, response_metadata={}, id='3962d25f-21a3-48d9-8fca-c864aaf6389f'), AIMessage(content='<|start|>assistant<|channel|>commentary to=functions.weather-get-weather <|constrain|>json<|message|>{\n  "location": "New York City"\n}<|call|>', additional_kwargs={}, response_metadata={'token_usage': {'prompt_tokens': 127, 'total_tokens': 189, 'completion_tokens': 62, 'completion_tokens_details': None, 'prompt_tokens_details': None}, 'model': 'Turbo', 'finish_reason': 'FinishReason.tool_calls'}, id='lc_run--05316aa0-997d-45c6-818e-f08a83c988b8-0', usage_metadata={'input_tokens': 127, 'output_tokens': 62, 'total_tokens': 189})]}

At the same time, if I directly use the FastMCP URL: http://{my_ip}:8010/mcp, I will get the following output:

{'messages': [HumanMessage(content='what is the weather in nyc?', additional_kwargs={}, response_metadata={}, id='2326dd2d-f265-49cd-b54a-2772b7a2c04e'), AIMessage(content='', additional_kwargs={}, response_metadata={'token_usage': {'prompt_tokens': 126, 'total_tokens': 193, 'completion_tokens': 67, 'completion_tokens_details': None, 'prompt_tokens_details': None}, 'model': 'Turbo', 'finish_reason': 'FinishReason.tool_calls'}, id='lc_run--9dc34bf9-87c1-459f-9f5b-7e720088943f-0', tool_calls=[{'name': 'get_weather', 'args': {'location': 'New York City'}, 'id': 'call_c895fc2f70bd4795af8b0e5c', 'type': 'tool_call'}], usage_metadata={'input_tokens': 126, 'output_tokens': 67, 'total_tokens': 193}), ToolMessage(content='{\n  "result": "It\'s always sunny in New York"\n}', name='get_weather', id='287dcb07-11fb-4a54-9c58-1ab384e262d3', tool_call_id='call_c895fc2f70bd4795af8b0e5c'), AIMessage(content='The current weather in New\u202fYork City is: **“It’s always sunny in New\u202fYork.”** 🌞', additional_kwargs={}, response_metadata={'token_usage': {'prompt_tokens': 176, 'total_tokens': 247, 'completion_tokens': 71, 'completion_tokens_details': None, 'prompt_tokens_details': None}, 'model': 'Turbo', 'finish_reason': 'FinishReason.stop'}, id='lc_run--2f28eaa1-896e-4074-96c4-42b5f64e0d19-0', usage_metadata={'input_tokens': 176, 'output_tokens': 71, 'total_tokens': 247})]}

As you can see, using the original MCP URL can successfully call the tool, but using the Virtual Server MCP URL it does not work. Please help resolve this issue.

---

🤔 Expected Behavior

What should have happened instead?

LangChain should be able to call tools normally via the Virtual Server MCP URL

---

📓 Logs / Error Output

Paste any relevant stack traces or logs here.
⚠️ Do not paste secrets, credentials, or tokens.

---

🧠 Environment Info

You can retrieve most of this from the /version endpoint.

Key	Value
Version or commit	v0.9.0 and v0.8.0
Runtime	Python 3.12
Platform / OS	windows
Container	Docker

---

🧩 Additional Context (optional)

Add any configuration details, flags, or related issues.

• gateway docker run with --network=host
• I can call tool on the web by using MCP Servers > Test