mcp-use compatibility

[LenaWirtz]

Hi dear development team!
we tried to use the Unity MCP Server together with mcp-use and got the error message:
2026-01-27 16:15:21,096 - mcp_use - ERROR - ❌ Error running query: 'position'
2026-01-27 16:15:21,097 - mcp_use - ERROR - ❌ Error during agent execution: 'position'
KeyError: 'position'

Here is the very simple Test script that we used: test.py
Might be an issue with mcp-use compatibility because with the openai-agents pagage we managed the server to run... Would it be possible to fix this?

[CoderGamester]

Hello @LenaWirtz

Thank you for your report.
'position' is not a tool or a resource on the MCP so is expected not to work.
Have you tried to invoke one of the other tools?
What prompts are you trying to use with the MCP?

[LenaWirtz]

Thank you for your reply.
I have no idea why it is using something like position, as my prompt is only to add a cube to the scene.

import asyncio
import os
from dotenv import load_dotenv
from langchain_openai import AzureChatOpenAI
from mcp_use import MCPAgent, MCPClient
import yaml

async def main():
    # Load environment variables
    load_dotenv()

    # Create configuration dictionary
    config = {
        "mcpServers": {
            "mcp-unity": {
            "command": "node",
            "args": [
                "C:/Users/lwirtz/Documents/Develop/Unity/MCPTest/Library/PackageCache/com.gamelovers.mcp-unity@4d59d9f5d3ad/Server~/build/index.js"
                ]
            }
        }
    }

    # Create MCPClient from configuration dictionary
    client = MCPClient.from_dict(config)

    # Create LLM
    config_path = os.getenv("CONFIG_FILE", "config.yml")
    with open(config_path, "r") as f:
        llm_config = yaml.safe_load(f)
    llm = AzureChatOpenAI(
        model=llm_config["llm"]["model"],
        azure_endpoint=llm_config["llm"]["base_url"],
        api_key=os.getenv(llm_config["llm"]["key_env_name"]),
        api_version=llm_config["llm"]["api_version"],
    )

    # Create agent with the client
    agent = MCPAgent(llm=llm, client=client, max_steps=30)

    # Run the query
    result = await agent.run(
        "Add a Cube to the scene",
    )
    print(f"\nResult: {result}")

if __name__ == "__main__":
    asyncio.run(main())

it might be used in the mcp_use initialization, with other MCP Servers it worked out of the box.

[CoderGamester]

You’re using the MCP via your own MCP Client. Have you tried using other clients, like Claude Code, Codex or Cursor? Sent from Outlook for Mac From: Lena Wirtz ***@***.***> Date: Wednesday, 4 February 2026 at 09:07 To: CoderGamester/mcp-unity ***@***.***> Cc: Miguel Tomas ***@***.***>, Comment ***@***.***> Subject: Re: [CoderGamester/mcp-unity] mcp-use compatibility (Issue #105) [https://avatars.githubusercontent.com/u/74858037?s=20&v=4]LenaWirtz left a comment (CoderGamester/mcp-unity#105)<#105 (comment)> Thank you for your reply. I have no idea why it is using something like position, as my prompt is only to add a cube to the scene. import asyncio import os from dotenv import load_dotenv from langchain_openai import AzureChatOpenAI from mcp_use import MCPAgent, MCPClient import yaml async def main(): # Load environment variables load_dotenv() # Create configuration dictionary config = { "mcpServers": { "mcp-unity": { "command": "node", "args": [ ***@***.***/Server~/build/index.js" ] } } } # Create MCPClient from configuration dictionary client = MCPClient.from_dict(config) # Create LLM config_path = os.getenv("CONFIG_FILE", "config.yml") with open(config_path, "r") as f: llm_config = yaml.safe_load(f) llm = AzureChatOpenAI( model=llm_config["llm"]["model"], azure_endpoint=llm_config["llm"]["base_url"], api_key=os.getenv(llm_config["llm"]["key_env_name"]), api_version=llm_config["llm"]["api_version"], ) # Create agent with the client agent = MCPAgent(llm=llm, client=client, max_steps=30) # Run the query result = await agent.run( "Add a Cube to the scene", ) print(f"\nResult: {result}") if __name__ == "__main__": asyncio.run(main()) it might be used in the mcp_use initialization, with other MCP Servers it worked out of the box. — Reply to this email directly, view it on GitHub<#105 (comment)>, or unsubscribe<https://github.com/notifications/unsubscribe-auth/ABO6EKRDIQ7DXRNKPE4FPLT4KGZFFAVCNFSM6AAAAACTBV5PS6VHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZTQNBWGIYTCMRYHE>. You are receiving this because you commented.

[LenaWirtz]

The client comes from the mcp-use python package: https://pypi.org/project/mcp-use/
With other servers it worked out without problems so i was wondering if it was an incompatibility with the mcp server definition.

[CoderGamester]

I see, I never used with that package. Are you able to push a fix? I can approve it and merge to main if it works Sent from Outlook for Mac From: Lena Wirtz ***@***.***> Date: Wednesday, 4 February 2026 at 09:58 To: CoderGamester/mcp-unity ***@***.***> Cc: Miguel Tomas ***@***.***>, Comment ***@***.***> Subject: Re: [CoderGamester/mcp-unity] mcp-use compatibility (Issue #105) [https://avatars.githubusercontent.com/u/74858037?s=20&v=4]LenaWirtz left a comment (CoderGamester/mcp-unity#105)<#105 (comment)> The client comes from the mcp-use python package: https://pypi.org/project/mcp-use/ With other servers it worked out without problems so i was wondering if it was an incompatibility with the mcp server definition. — Reply to this email directly, view it on GitHub<#105 (comment)>, or unsubscribe<https://github.com/notifications/unsubscribe-auth/ABO6EKTDKS676QGZW4IT4B34KG7EPAVCNFSM6AAAAACTBV5PS6VHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZTQNBWGQ2TIOJTG4>. You are receiving this because you commented.

[LenaWirtz]

I am unfamiliar with the server implementation... so no idea :D
I could open a question in the mcp_use package.