Passing A URL Parameter to SSE Transport

[maxxrdrgz]

Hello,

I'm trying to setup this MCP server as a remote server using SSE Transport. I'm trying to gauge how feasible it might be for me to modify the code to receive a url parameter from the client, and using that in the JiraFetcher (or the JiraClient) object. For example, in the server.py code, you can certainly parse URL parameters like my example below. But I'm having trouble passing the parameter elsewhere in the code. I'd be happy to contribute if someone else can nudge me in the right directions (or let me know if this is not possible).

The client's url might look like this:
"url": "http://SERVER:8000/sse?jira_api_token=JIRAPERSONALTOKEN",

async def run_server(transport: str = "stdio", port: int = 8000) -> None:
    """Run the MCP Atlassian server with the specified transport."""
    if transport == "sse":
        from mcp.server.sse import SseServerTransport
        from starlette.applications import Starlette
        from starlette.requests import Request
        from starlette.routing import Mount, Route

        sse = SseServerTransport("/messages/")

        async def handle_sse(request: Request) -> None:
            # Extract query parameters
            query_params = request.query_params
            api_key = query_params.get("jira_api_token")  # Get the API key from query parameters

            if not api_key:
                print("API key is missing!")
                return

            print(f"Received API Key: {api_key}")

            async with sse.connect_sse(
                request.scope, request.receive, request._send
            ) as streams:
                await app.run(
                    streams[0], streams[1], app.create_initialization_options()
                )

[maxxrdrgz]

@sooperset , you're the creator right? If you have some time to nudge me in the right direction I'd be happy to help :-).

I'm more of a javascript guy and so I'm having issues passing the parsed URL parameter outside of the run_server() and into something like get_available_services().

[sooperset]

Hey @maxxrdrgz! I think this should be possible, though I haven't tested it myself. You'd need to:

1. In server.py, try modifying handle_sse() to add your parameter to the initialization options before passing to app.run()

2. Then see if you can access those parameters in server_lifespan() through server.initialization_options when creating the JiraFetcher

The Server class from MCP might pass those initialization options through to your lifespan function, but you'll need to test if this approach works. If it doesn't, there may be other ways to pass context between these components.

One security consideration - URL params can be visible in logs and browser history, so keep that in mind if you're passing sensitive tokens this way.

[maxxrdrgz]

Thanks. I thought about the security concern however, it's not really any different if you're using the JIRA rest api with bearer authentication over https.

Thank you. I shall give your advice a try over the next few days.

[maxxrdrgz]

If anyone stumbles upon this issue, I managed to allow multi-user via my PR here: #224

[sooperset]

Closing this issue as it relates to the broader goal of supporting multiple users/instances with dynamic configurations, which is now being tracked centrally in #231.

Please follow #231 for future updates. Thanks for raising this aspect!