Lad

If you're building AI agents that need to operate on local networks — hotel Wi-Fi, office LANs, cruise ships — LAD-A2A fills a hole that nothing else addresses directly. You could hand-roll mDNS discovery, but you'd probably make the mistakes this protocol already solves, especially around trust. The separation of channel authentication from identity verification is the kind of detail that saves you from a security incident later. We'd reach for this when you need zero-config discovery in a heterogeneous environment. The spec is designed to hand off to standard A2A for communication, so it fits naturally into an existing agent architecture. The reference implementation in Python gives you a concrete starting point, and the interactive demo is a fast way to see the whole flow. Where it bites: this is not a product. There's no managed service, no UI for end users, no plugin for your favorite assistant. You're integrating a spec into your own codebase. Production requires TLS 1.2+ enforcement — the client refuses non-https URLs by default, which is correct but means you must configure certificates on every local agent. That's real work. The closest alternative is rolling your own discovery with mDNS or a centralized registry. Centralized registries scale across WAN, but they break on isolated networks and add a dependency. LAD-A2A is decentralized and built for edge cases. For non-technical users expecting a turnkey platform, it's not for you — yet. In practice, we'd pair LAD-A2A with A2A for communication and MCP for tools, and we'd budget time for the integration. The trust model is the reason to pick it over a quick script; it's the difference between 'it works on my machine' and 'it works on a hostile network.'