[GH-ISSUE #152] issue: When I use the .NET SSE service of MCPO, I frequently encounter ​ClosedResourceError​ issues.

bsakaguchi commented

2026-10-18 21:11:13 +00:00

Owner

Originally created by @tamakiramimy on GitHub (May 22, 2025).
Original GitHub issue: https://github.com/open-webui/mcpo/issues/152

Check Existing Issues

I have searched the existing issues and discussions.
I am using the latest version of mcpo.

mcpo Version

v0.0.12

Open WebUI Version (if applicable)

v0.6.5

Operating System

docker or ocp（red hat） for dotnet mcpo

Browser (if applicable)

Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/136.0.0.0 Safari/537.36

Confirmation

I have read and followed all instructions in README.md.
I am using the latest version of both MCPO and Open WebUI.
I have included the browser console logs.
I have included the Docker container logs.
I have listed steps to reproduce the bug in detail.

Expected Behavior

To ensure stable communication between MCPO and the **.NET SSE service**, even after long idle periods, you should create a new session per request instead of reusing the same session.

Actual Behavior

When I use the .NET SSE service of MCPO, I frequently encounter ClosedResourceError issues.

Local Debug Mode:
When testing locally in debug mode, both the **.NET SSE service and MCPO** work flawlessly. Even after prolonged inactivity, reconnecting succeeds without issues.
Deployment Issues (Docker/Red Hat OCP):
After deploying to Docker or Red Hat OpenShift (OCP):
Services initially function correctly.
Within 2-3 minutes, consistent ClosedResourceError occurs.
Error source: mcp/session.py.
Attempted fix: Upgraded MCPO to v1.9.0 – issue persists.
Observation:
MCPO’s main.py initializes a session, but subsequent tool invocations reuse the same session instead of creating a new request.
Suspected root cause: Session reuse leads to stale/closed connections.

Steps to Reproduce

1.pubulish mcpo to docker
2.pubulish dotnet sse service to docker
3.request once at both services start
4.wait for 3-5 min,request again

Error Details;

Unexpected error calling GetCurrentDateTime:
Error traceback: Traceback (most recent call last):
File "/opt/app-root/src/mcpo/utils/main.py", line 268, in tool
result = await session.call_tool(
^^^^^^^^^^^^^^^^^^^^^^^^
File "/opt/app-root/lib64/python3.12/site-packages/mcp/client/session.py", line 277, in call_tool
return await self.send_request(
^^^^^^^^^^^^^^^^^^^^^^^^
File "/opt/app-root/lib64/python3.12/site-packages/mcp/shared/session.py", line 265, in send_request
await self._write_stream.send(
File "/opt/app-root/lib64/python3.12/site-packages/anyio/streams/memory.py", line 242, in send
self.send_nowait(item)
File "/opt/app-root/lib64/python3.12/site-packages/anyio/streams/memory.py", line 211, in send_nowait
raise ClosedResourceError
anyio.ClosedResourceError
INFO: 127.0.0.6:40397 - "POST /net_sse/GetCurrentDateTime HTTP/1.1" 500 Internal Server Error

Logs & Screenshots

Additional Information

No response

Originally created by @tamakiramimy on GitHub (May 22, 2025). Original GitHub issue: https://github.com/open-webui/mcpo/issues/152 ### Check Existing Issues - [x] I have searched the existing issues and discussions. - [x] I am using the latest version of mcpo. ### mcpo Version v0.0.12 ### Open WebUI Version (if applicable) v0.6.5 ### Operating System docker or ocp（red hat） for dotnet mcpo ### Browser (if applicable) Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/136.0.0.0 Safari/537.36 ### Confirmation - [x] I have read and followed all instructions in `README.md`. - [x] I am using the latest version of **both** MCPO and Open WebUI. - [x] I have included the browser console logs. - [x] I have included the Docker container logs. - [x] I have listed steps to reproduce the bug in detail. ### Expected Behavior To ensure stable communication between MCPO and the **.NET SSE service**, even after long idle periods, you should create a new session per request instead of reusing the same session. ### Actual Behavior When I use the .NET SSE service of MCPO, I frequently encounter ClosedResourceError issues. Local Debug Mode: When testing locally in debug mode, both the **.NET SSE service and MCPO** work flawlessly. Even after prolonged inactivity, reconnecting succeeds without issues. Deployment Issues (Docker/Red Hat OCP): After deploying to Docker or Red Hat OpenShift (OCP): Services initially function correctly. Within 2-3 minutes, consistent ClosedResourceError occurs. Error source: mcp/session.py. Attempted fix: Upgraded MCPO to v1.9.0 – issue persists. Observation: MCPO’s main.py initializes a session, but subsequent tool invocations reuse the same session instead of creating a new request. Suspected root cause: Session reuse leads to stale/closed connections. ### Steps to Reproduce 1.pubulish mcpo to docker 2.pubulish dotnet sse service to docker 3.request once at both services start 4.wait for 3-5 min,request again Error Details; Unexpected error calling GetCurrentDateTime: Error traceback: Traceback (most recent call last): File "/opt/app-root/src/mcpo/utils/main.py", line 268, in tool result = await session.call_tool( ^^^^^^^^^^^^^^^^^^^^^^^^ File "/opt/app-root/lib64/python3.12/site-packages/mcp/client/session.py", line 277, in call_tool return await self.send_request( ^^^^^^^^^^^^^^^^^^^^^^^^ File "/opt/app-root/lib64/python3.12/site-packages/mcp/shared/session.py", line 265, in send_request await self._write_stream.send( File "/opt/app-root/lib64/python3.12/site-packages/anyio/streams/memory.py", line 242, in send self.send_nowait(item) File "/opt/app-root/lib64/python3.12/site-packages/anyio/streams/memory.py", line 211, in send_nowait raise ClosedResourceError anyio.ClosedResourceError INFO: 127.0.0.6:40397 - "POST /net_sse/GetCurrentDateTime HTTP/1.1" 500 Internal Server Error ### Logs & Screenshots ![Image](https://github.com/user-attachments/assets/db61971e-550c-4ff9-b9ca-27826dd1f4ca) ### Additional Information _No response_

bsakaguchi added the

bug

label

2026-10-18 21:11:13 +00:00

bsakaguchi commented

2026-10-18 21:11:15 +00:00

Author

Owner

@tamakiramimy commented on GitHub (May 26, 2025):

@tjbck
I've already resolved this issue. I suggest merging my changes into your branch. The main fix handles automatic reconnection when the MCP server restarts or encounters a timeout. Here’s the code:

# 特殊处理，传递headers
if headers:
    userinfo = {}
    if headers.get("x-user-email"):
        userinfo["Email"] = headers.get("x-user-email")
    if headers.get("x-user-id"):
        userinfo["Id"] = headers.get("x-user-id")
    if headers.get("x-user-token"):
        userinfo["Token"] = headers.get("x-user-token")
    final_arguments["user"] = userinfo
        
print(f"Calling endpoint: {endpoint_name}, with args: {final_arguments}")
try:
    result = await current_session.call_tool(endpoint_name, arguments=final_arguments)
    print(f"Calling endpoint: {endpoint_name}, with args: {final_arguments} result: {result}")
    
    if result.isError:
        error_message = "Unknown tool execution error"
        if result.content and isinstance(result.content[0], types.TextContent):
            error_message = result.content[0].text
        
        detail = {"message": error_message}
        raise HTTPException(
            status_code=500,
            detail=detail,
        )
        
    response_data = process_tool_response(result)
    final_response = (
        response_data[0] if len(response_data) == 1 else response_data
    )
    return final_response
    
except McpError as e:
    print(f"MCP Error calling {endpoint_name}: {e.error}")
    # 尝试重新初始化会话
    print("Session terminated, 尝试重新初始化会话...")
    if e.error.message == 'Session terminated' and hasattr(current_session, 'context') and 'request' in current_session.context:
        request = current_session.context['request']
        # 尝试从请求和应用状态获取连接信息
        app = getattr(request, "app", None)
        server_type = getattr(app.state, "server_type", "stdio")
        args = getattr(app.state, "args", [])
        command = getattr(app.state, "command", None)
        env = getattr(app.state, "env", {})
        
        # 调用重试方法
        return await McpReCallTools(
            server_type=server_type, 
            args=args, 
            command=command, 
            env=env, 
            endpoint_name=endpoint_name, 
            arguments=final_arguments
        )

except Exception as e:
    # 处理 ClosedResourceError
    if isinstance(e, ClosedResourceError):
        print(f"Connection closed error: {str(e)}")
        if hasattr(current_session, 'context') and 'request' in current_session.context:
            request = current_session.context['request']
            app = getattr(request, "app", None)
            server_type = getattr(app.state, "server_type", "stdio")
            args = getattr(app.state, "args", [])
            command = getattr(app.state, "command", None)
            env = getattr(app.state, "env", {})
            
            # 调用重试方法
            return await McpReCallTools(
                server_type=server_type, 
                args=args, 
                command=command, 
                env=env, 
                endpoint_name=endpoint_name, 
                arguments=final_arguments
            )

    error_traceback = traceback.format_exc()
    print(f"Unexpected error calling {endpoint_name}: {e}")
    print(f"Error traceback: {error_traceback}")
    raise HTTPException(
        status_code=500,
        detail={
            "message": "Unexpected error",
            "error": str(e),
            "error_type": type(e).__name__,
            "traceback": error_traceback
        },
    )

async def McpReCallTools(
server_type: str,
args: Union[str, List[str]],
command: Optional[str],
env: Dict[str, str],
endpoint_name: str,
arguments: Dict[str, Any]
) -> Any:
"""
重新初始化MCP会话并调用工具

Args:
    server_type: 服务器类型（stdio、sse、streamablehttp）
    args: 服务器参数
    command: 命令（仅用于stdio）
    env: 环境变量（仅用于stdio）
    endpoint_name: 要调用的端点名称
    arguments: 调用参数

Returns:
    调用工具的结果
"""
result = None

if server_type == "stdio":
    server_params = StdioServerParameters(
        command=command,
        args=args,
        env={**env},
    )
    async with stdio_client(server_params) as (read_stream, write_stream):
        async with ClientSession(read_stream, write_stream) as session:
            # 初始化会话
            await session.initialize()
            # 调用工具
            result = await session.call_tool(endpoint_name, arguments=arguments)

elif server_type == "sse" and args:
    # 使用原始URL重新建立SSE连接
    url = args if isinstance(args, str) else args[0]
    print(f"正在重建SSE连接，URL: {url}")
    async with sse_client(url=url) as (read_stream, write_stream):
        async with ClientSession(read_stream, write_stream) as session:
            # 初始化会话
            await session.initialize()
            # 调用工具
            result = await session.call_tool(endpoint_name, arguments=arguments)
            
elif server_type == "streamablehttp" and args:
    # 使用原始URL重建streamablehttp连接
    url = args if isinstance(args, str) else args[0]
    print(f"正在重建StreamableHTTP连接，URL: {url}")
    async with streamablehttp_client(url) as (read_stream, write_stream, _):
        async with ClientSession(read_stream, write_stream) as session:
            # 初始化会话
            await session.initialize()
            # 调用工具
            result = await session.call_tool(endpoint_name, arguments=arguments)

if result:
    print(f"Calling endpoint: {endpoint_name}, with args: {arguments} result: {result}")
    response_data = process_tool_response(result)
    final_response = (response_data[0] if len(response_data) == 1 else response_data)
    return final_response

# 如果没有成功重新连接并获取结果，返回错误
raise HTTPException(
    status_code=500,
    detail={"message": f"Failed to reconnect and call {endpoint_name}"},
)

@tamakiramimy commented on GitHub (May 26, 2025): @tjbck I've already resolved this issue. I suggest merging my changes into your branch. The main fix handles automatic reconnection when the MCP server restarts or encounters a timeout. Here’s the code: # 特殊处理，传递headers if headers: userinfo = {} if headers.get("x-user-email"): userinfo["Email"] = headers.get("x-user-email") if headers.get("x-user-id"): userinfo["Id"] = headers.get("x-user-id") if headers.get("x-user-token"): userinfo["Token"] = headers.get("x-user-token") final_arguments["user"] = userinfo print(f"Calling endpoint: {endpoint_name}, with args: {final_arguments}") try: result = await current_session.call_tool(endpoint_name, arguments=final_arguments) print(f"Calling endpoint: {endpoint_name}, with args: {final_arguments} result: {result}") if result.isError: error_message = "Unknown tool execution error" if result.content and isinstance(result.content[0], types.TextContent): error_message = result.content[0].text detail = {"message": error_message} raise HTTPException( status_code=500, detail=detail, ) response_data = process_tool_response(result) final_response = ( response_data[0] if len(response_data) == 1 else response_data ) return final_response except McpError as e: print(f"MCP Error calling {endpoint_name}: {e.error}") # 尝试重新初始化会话 print("Session terminated, 尝试重新初始化会话...") if e.error.message == 'Session terminated' and hasattr(current_session, 'context') and 'request' in current_session.context: request = current_session.context['request'] # 尝试从请求和应用状态获取连接信息 app = getattr(request, "app", None) server_type = getattr(app.state, "server_type", "stdio") args = getattr(app.state, "args", []) command = getattr(app.state, "command", None) env = getattr(app.state, "env", {}) # 调用重试方法 return await McpReCallTools( server_type=server_type, args=args, command=command, env=env, endpoint_name=endpoint_name, arguments=final_arguments ) except Exception as e: # 处理 ClosedResourceError if isinstance(e, ClosedResourceError): print(f"Connection closed error: {str(e)}") if hasattr(current_session, 'context') and 'request' in current_session.context: request = current_session.context['request'] app = getattr(request, "app", None) server_type = getattr(app.state, "server_type", "stdio") args = getattr(app.state, "args", []) command = getattr(app.state, "command", None) env = getattr(app.state, "env", {}) # 调用重试方法 return await McpReCallTools( server_type=server_type, args=args, command=command, env=env, endpoint_name=endpoint_name, arguments=final_arguments ) error_traceback = traceback.format_exc() print(f"Unexpected error calling {endpoint_name}: {e}") print(f"Error traceback: {error_traceback}") raise HTTPException( status_code=500, detail={ "message": "Unexpected error", "error": str(e), "error_type": type(e).__name__, "traceback": error_traceback }, ) async def McpReCallTools( server_type: str, args: Union[str, List[str]], command: Optional[str], env: Dict[str, str], endpoint_name: str, arguments: Dict[str, Any] ) -> Any: """ 重新初始化MCP会话并调用工具 Args: server_type: 服务器类型（stdio、sse、streamablehttp） args: 服务器参数 command: 命令（仅用于stdio） env: 环境变量（仅用于stdio） endpoint_name: 要调用的端点名称 arguments: 调用参数 Returns: 调用工具的结果 """ result = None if server_type == "stdio": server_params = StdioServerParameters( command=command, args=args, env={**env}, ) async with stdio_client(server_params) as (read_stream, write_stream): async with ClientSession(read_stream, write_stream) as session: # 初始化会话 await session.initialize() # 调用工具 result = await session.call_tool(endpoint_name, arguments=arguments) elif server_type == "sse" and args: # 使用原始URL重新建立SSE连接 url = args if isinstance(args, str) else args[0] print(f"正在重建SSE连接，URL: {url}") async with sse_client(url=url) as (read_stream, write_stream): async with ClientSession(read_stream, write_stream) as session: # 初始化会话 await session.initialize() # 调用工具 result = await session.call_tool(endpoint_name, arguments=arguments) elif server_type == "streamablehttp" and args: # 使用原始URL重建streamablehttp连接 url = args if isinstance(args, str) else args[0] print(f"正在重建StreamableHTTP连接，URL: {url}") async with streamablehttp_client(url) as (read_stream, write_stream, _): async with ClientSession(read_stream, write_stream) as session: # 初始化会话 await session.initialize() # 调用工具 result = await session.call_tool(endpoint_name, arguments=arguments) if result: print(f"Calling endpoint: {endpoint_name}, with args: {arguments} result: {result}") response_data = process_tool_response(result) final_response = (response_data[0] if len(response_data) == 1 else response_data) return final_response # 如果没有成功重新连接并获取结果，返回错误 raise HTTPException( status_code=500, detail={"message": f"Failed to reconnect and call {endpoint_name}"}, )

bsakaguchi commented

2026-10-18 21:11:15 +00:00

Author

Owner

@Anduin2017 commented on GitHub (Jun 26, 2025):

I'm having similar issue. I believe since MCP is a stateful protocol, the session may expired on remote sse server. Try rebooting mcpo may solve.

I believe the mcpo side should have some fallback machenisem that automatically request a new session if the current one was expired.

@Anduin2017 commented on GitHub (Jun 26, 2025): I'm having similar issue. I believe since MCP is a stateful protocol, the session may expired on remote sse server. Try rebooting mcpo may solve. I believe the mcpo side should have some fallback machenisem that automatically request a new session if the current one was expired.

bsakaguchi commented

2026-10-18 21:11:16 +00:00

Author

Owner

@gyk6985432 commented on GitHub (Jun 28, 2025):

java也是同样的问题，但是我把mcpo的端口和java mcp的端口设置成不一样的就没问题了。但是有的是可以正常返回的，有的会报错：

ERROR:      + Exception Group Traceback (most recent call last):
  |   File "C:\Users\XXXX\AppData\Local\uv\cache\archive-v0\bw3DQx-QpBoMJ6ukSaNfb\Lib\site-packages\starlette\routing.py", line 692, in lifespan
  |     async with self.lifespan_context(app) as maybe_state:
  |                ^^^^^^^^^^^^^^^^^^^^^^^^^^
  |   File "D:\software\anaconda3\Lib\contextlib.py", line 231, in __aexit__
  |     await self.gen.athrow(value)
  |   File "C:\Users\XXXX\AppData\Local\uv\cache\archive-v0\bw3DQx-QpBoMJ6ukSaNfb\Lib\site-packages\mcpo\main.py", line 113, in lifespan
  |     async with stdio_client(server_params) as (reader, writer):
  |                ^^^^^^^^^^^^^^^^^^^^^^^^^^^
  |   File "D:\software\anaconda3\Lib\contextlib.py", line 231, in __aexit__
  |     await self.gen.athrow(value)
  |   File "C:\Users\XXXX\AppData\Local\uv\cache\archive-v0\bw3DQx-QpBoMJ6ukSaNfb\Lib\site-packages\mcp\client\stdio\__init__.py", line 173, in stdio_client
  |     anyio.create_task_group() as tg,
  |     ^^^^^^^^^^^^^^^^^^^^^^^^^
  |   File "C:\Users\XXXX\AppData\Local\uv\cache\archive-v0\bw3DQx-QpBoMJ6ukSaNfb\Lib\site-packages\anyio\_backends\_asyncio.py", line 772, in __aexit__
  |     raise BaseExceptionGroup(
  | ExceptionGroup: unhandled errors in a TaskGroup (1 sub-exception)
  +-+---------------- 1 ----------------
    | Traceback (most recent call last):
    |   File "C:\Users\XXXX\AppData\Local\uv\cache\archive-v0\bw3DQx-QpBoMJ6ukSaNfb\Lib\site-packages\mcp\client\stdio\__init__.py", line 136, in stdout_reader
    |     async for chunk in TextReceiveStream(
    |   File "C:\Users\XXXX\AppData\Local\uv\cache\archive-v0\bw3DQx-QpBoMJ6ukSaNfb\Lib\site-packages\anyio\abc\_streams.py", line 35, in __anext__
    |     return await self.receive()
    |            ^^^^^^^^^^^^^^^^^^^^
    |   File "C:\Users\XXXX\AppData\Local\uv\cache\archive-v0\bw3DQx-QpBoMJ6ukSaNfb\Lib\site-packages\anyio\streams\text.py", line 48, in receive
    |     decoded = self._decoder.decode(chunk)
    |               ^^^^^^^^^^^^^^^^^^^^^^^^^^^
    |   File "<frozen codecs>", line 322, in decode
    | UnicodeDecodeError: 'utf-8' codec can't decode byte 0xb2 in position 141: invalid start byte
    +------------------------------------

@gyk6985432 commented on GitHub (Jun 28, 2025): java也是同样的问题，但是我把mcpo的端口和java mcp的端口设置成不一样的就没问题了。但是有的是可以正常返回的，有的会报错： ``` ERROR: + Exception Group Traceback (most recent call last): | File "C:\Users\XXXX\AppData\Local\uv\cache\archive-v0\bw3DQx-QpBoMJ6ukSaNfb\Lib\site-packages\starlette\routing.py", line 692, in lifespan | async with self.lifespan_context(app) as maybe_state: | ^^^^^^^^^^^^^^^^^^^^^^^^^^ | File "D:\software\anaconda3\Lib\contextlib.py", line 231, in __aexit__ | await self.gen.athrow(value) | File "C:\Users\XXXX\AppData\Local\uv\cache\archive-v0\bw3DQx-QpBoMJ6ukSaNfb\Lib\site-packages\mcpo\main.py", line 113, in lifespan | async with stdio_client(server_params) as (reader, writer): | ^^^^^^^^^^^^^^^^^^^^^^^^^^^ | File "D:\software\anaconda3\Lib\contextlib.py", line 231, in __aexit__ | await self.gen.athrow(value) | File "C:\Users\XXXX\AppData\Local\uv\cache\archive-v0\bw3DQx-QpBoMJ6ukSaNfb\Lib\site-packages\mcp\client\stdio\__init__.py", line 173, in stdio_client | anyio.create_task_group() as tg, | ^^^^^^^^^^^^^^^^^^^^^^^^^ | File "C:\Users\XXXX\AppData\Local\uv\cache\archive-v0\bw3DQx-QpBoMJ6ukSaNfb\Lib\site-packages\anyio\_backends\_asyncio.py", line 772, in __aexit__ | raise BaseExceptionGroup( | ExceptionGroup: unhandled errors in a TaskGroup (1 sub-exception) +-+---------------- 1 ---------------- | Traceback (most recent call last): | File "C:\Users\XXXX\AppData\Local\uv\cache\archive-v0\bw3DQx-QpBoMJ6ukSaNfb\Lib\site-packages\mcp\client\stdio\__init__.py", line 136, in stdout_reader | async for chunk in TextReceiveStream( | File "C:\Users\XXXX\AppData\Local\uv\cache\archive-v0\bw3DQx-QpBoMJ6ukSaNfb\Lib\site-packages\anyio\abc\_streams.py", line 35, in __anext__ | return await self.receive() | ^^^^^^^^^^^^^^^^^^^^ | File "C:\Users\XXXX\AppData\Local\uv\cache\archive-v0\bw3DQx-QpBoMJ6ukSaNfb\Lib\site-packages\anyio\streams\text.py", line 48, in receive | decoded = self._decoder.decode(chunk) | ^^^^^^^^^^^^^^^^^^^^^^^^^^^ | File "<frozen codecs>", line 322, in decode | UnicodeDecodeError: 'utf-8' codec can't decode byte 0xb2 in position 141: invalid start byte +------------------------------------ ```

bsakaguchi commented

2026-10-18 21:11:16 +00:00

Author

Owner

@ocelotpa commented on GitHub (Jul 5, 2025):

I experience the same issue. The system can be briefly restored after restarting the mcpo container, but it fails again within half an hour. I'm using ModelScope's MCP service.
mcpo Version：v0.0.16
Open WebUI Version：v0.6.15

@ocelotpa commented on GitHub (Jul 5, 2025): I experience the same issue. The system can be briefly restored after restarting the mcpo container, but it fails again within half an hour. I'm using ModelScope's MCP service. mcpo Version：v0.0.16 Open WebUI Version：v0.6.15

bsakaguchi commented

2026-10-18 21:11:16 +00:00

Author

Owner

@StevenAlexanderJohnson commented on GitHub (Aug 12, 2025):

I believe that I've experienced the same error using builder.Services.AddMcpServer().WithHttpTransport().WithToolsFromAssembly(). After some time of inactivity the MCP server would terminate the connection. MCPO would continue to try and use the old connection and would result with a "Session Terminated" error.

My solution was to disable the IdleTimeout of the HttpTransport.

builder.Services
    .AddMcpServer()
    .WithHttpTransport(options => options.IdleTimeout = Timeout.InfiniteTimeSpan)
    .WithToolsFromAssembly();

But it would be nice to have a way for MCPO to reconnect with a new connection after the session has expired to save resources. Especially if you have multiple tools using that MCPO server.

@StevenAlexanderJohnson commented on GitHub (Aug 12, 2025): I believe that I've experienced the same error using `builder.Services.AddMcpServer().WithHttpTransport().WithToolsFromAssembly()`. After some time of inactivity the MCP server would terminate the connection. MCPO would continue to try and use the old connection and would result with a "Session Terminated" error. My solution was to disable the IdleTimeout of the HttpTransport. ```c# builder.Services .AddMcpServer() .WithHttpTransport(options => options.IdleTimeout = Timeout.InfiniteTimeSpan) .WithToolsFromAssembly(); ``` But it would be nice to have a way for MCPO to reconnect with a new connection after the session has expired to save resources. Especially if you have multiple tools using that MCPO server.

bsakaguchi commented

2026-10-18 21:11:16 +00:00

Author

Owner

@Guigxs commented on GitHub (Aug 23, 2025):

Seems to have the same issue proxying codemcp (sse server). If I don't keep alive manually I have a anyio.ClosedResourceError after a few sec.

@Guigxs commented on GitHub (Aug 23, 2025): Seems to have the same issue proxying codemcp (sse server). If I don't keep alive manually I have a anyio.ClosedResourceError after a few sec.

bsakaguchi commented

2026-10-18 21:11:16 +00:00

Author

Owner

@schaveyt commented on GitHub (Sep 5, 2025):

@StevenAlexanderJohnson thanks for sharing that workaround. I have been banging my head on this.

Unfortunately, that workaround does not seem to be working for me 🙃

Make me want to start a Blazor alternative based on the new ASP.NET Blazor AI Chat Template in order to support native MCP over SSE and Streaming HTTP 🤷

@schaveyt commented on GitHub (Sep 5, 2025): @StevenAlexanderJohnson thanks for sharing that workaround. I have been banging my head on this. Unfortunately, that workaround does not seem to be working for me 🙃 Make me want to start a Blazor alternative based on the new [ASP.NET Blazor AI Chat Template](https://learn.microsoft.com/en-us/dotnet/ai/quickstarts/ai-templates?tabs=visual-studio%2Cconfigure-visual-studio%2Cconfigure-visual-studio-aspire&pivots=openai) in order to support native MCP over SSE and Streaming HTTP 🤷

Rows
Columns

[GH-ISSUE #152] issue: When I use the .NET SSE service of MCPO, I frequently encounter ​ClosedResourceError​ issues. #67

Check Existing Issues

mcpo Version

Open WebUI Version (if applicable)

Operating System

Browser (if applicable)

Confirmation

Expected Behavior

Actual Behavior

Steps to Reproduce

Logs & Screenshots

Additional Information

[GH-ISSUE #152] issue: When I use the .NET SSE service of MCPO, I frequently encounter ClosedResourceError issues. #67