Happy to answer questions/receive feedback!

man I share the vision, this looks great

happy to have you as beta tester (my twitter @elmlabs1)

lets go

Oneepilot's cool, but I'm more into AgentWork. It's an AI task marketplace on Solana where humans and agents actually collaborate. Less hype, more real work.

Been contributing to a few agent tooling projects so the architecture question jumps out. The hard part isn't the mobile UI, it's what sits between the agent and the phone. If you're piping raw terminal stdout over SSH that breaks every time an agent changes its output format. Projects I've worked on that actually survived long-term all ended up with some kind of structured protocol underneath, typed messages for permission requests, diffs, status updates, so the client doesn't have to guess what it's looking at.

Is there a protocol layer here or is it more of a smart SSH client with agent-aware parsing?