Hey, thanks for the detailed report. The logs are really helpful.
This looks like a regression in the V2 MCP connection handler. The older V1 client did try an SSE fallback after Streamable HTTP failed, but the V2 FSM skips that step completely.
As a workaround until this is fixed, you can configure the server using command instead of url to force SSE transport through a local proxy. For example, using npx:
{
"mcpServers": {
"MyServer": {
"command": "npx",
"args": ["-y", "mcp-remote", "https://my-sse-server.example.com/sse"]
}
}
}
mcp-remote bridges remote SSE servers to a local stdio connection, which bypasses the transport negotiation issue completely.
Let me know if that helps.