Ask HN: Do you think a new alternative to MCP would be useful?
Hi everyone,
I've been wrestling with MCP, and keep finding the same weaknesses: - the middleware server feels completely unncessary and overly complex - it has too many security flaws due to what feels like a rushed out standard - it's just so poorly undocumented and confusing
I've been thinking of doing a new protocol, but am too aware of that ol' comic (https://imgs.xkcd.com/comics/standards_2x.png)
Yet was curious to ask you guys - is there value in trying to build a new scalable, secure version of MCP?
Or should I just stick to the protocol that's currently popular?
The proof of the pudding is in the eating. That is, if you can produce a working (free) implementation demonstrating the superiority of your protocol then there is huge value. If you can't there is no value.
JS was designed in a few days. And many years later, people made proposals to it and improved it. A few like TS built on top of it.
Standards that are well designed end up like Java - still flawed and clunky. And built on top as well.
MCP wasn't designed by anyone omniscient. So if you found flaws with it, go ahead and propose something.