I've been following the issue in #386 and see the solution merged in #540. However, it feels very wrong. An app should never have to know its mount path in advance in order to work properly. Furthermore, this results in a cumbersome DX where either users specify the mount paths when creating the FastMCP server, effectively preventing it from being used in any other context or users double-specify mount paths when mounting the SSE app into another Starlette application, which is confusing and awkward.

With that in mind, consider this example from the docs of #540. The GitHub and browser apps will simply break if they are mounted under any other path. This is not a good user-facing outcome.

# Create Starlette app with multiple mounted servers
app = Starlette(
    routes=[
        # Using settings-based configuration
        Mount("/github", app=github_mcp.sse_app()),
        Mount("/browser", app=browser_mcp.sse_app()),
        # Using direct mount path parameter
        Mount("/curl", app=curl_mcp.sse_app("/curl")),
        Mount("/search", app=search_mcp.sse_app("/search")),
    ]
)

Happily I don't think we need to settle, or pollute the FastMCP server with new user-facing configuration. The ASGI spec provides for pretty much this exact circumstance with a root_path attribute on the Scope object. By examining the root path when constructing the messages POST endpoint, we can ensure that it is always generated at the correct absolute path, without having to perform any complex parsing or even asking users to provide additional configuration.

With this PR, the following works out of the box:

import uvicorn
from starlette.applications import Starlette
from starlette.routing import Mount

from mcp.server.fastmcp import FastMCP

mcp = FastMCP()


@mcp.tool()
async def hello() -> str:
    return "Hello, world!"


sse_app = mcp.sse_app()
main_app = Starlette(routes=[Mount("/nested/mount/", app=sse_app)])

if __name__ == "__main__":
    uvicorn.run(main_app, host="127.0.0.1", port=8000)