Love the hosted SSE/HTTP mirrors and registry. We built something similar to catch handshake drift in our remote MCP clients, and being able to replay the JSON-RPC log in-browser saved us from chasing environment-specific timeouts because we could rerun the exact frame sequence from the failure. One idea: add a little health badge per server showing last-seen transport (HTTP/SSE/WS) plus a running error count so integrators know which endpoint matches their stack before they flip the switch.