Show HN: Middleware for translating between AI agent protocols

I’ve been working on a problem that kept coming up while building multi-agent systems: agents don’t interoperate cleanly.

Even when two agents can technically connect, they often: - use different schemas - structure intent differently - break when assumptions don’t match

So you end up writing a lot of brittle glue code just to pass messages around.

I built a middleware layer to sit between agents and handle this:

- schema-aware translation between protocols - semantic mapping (so intent stays consistent across formats) - orchestration of agent-to-agent handoffs

The goal is to let agents communicate without forcing everything into a single standard or rewriting existing systems.

One thing I’m still figuring out is how far abstraction should go — at some point, “translation” can hide important differences instead of resolving them.

Curious how others are thinking about this: - Are you standardizing on one protocol? - Building internal adapters? - Or avoiding cross-agent interoperability altogether?

Would really appreciate any thoughts or criticism.

Project: https://www.useengram.com/