The Agent Protocols Stopped at the Client — Let's Talk About the Server Half
Every few weeks another "skills" announcement lands. Prompt bundles, client-side toolkits, SDK-level orchestration patterns. I read them the way you probably do — genuinely interested, quietly uneasy. The industry's channel for developer agency in the agent stack has ended up almost entirely on the client side. Which is strange, if you think about where capabilities actually have to run: against…
Many recent "skills" announcements have highlighted the client-side focus in the agent stack. However, the critical capabilities required for agents are actually server-side. This post introduces an open-source Java library, Franca, to fill this gap. The OpenAI and Anthropic message formats have become the primary protocols for agents, offering impressive features on the client side, including conversation shape, tool declarations, reasoning controls, and streaming semantics.
A client SDK can communicate with any server that supports the same protocol. Yet, the server side of the equation remains undefined, leaving developers with limited options. The current landscape includes vendor platforms, conversion gateways, and self-hosted inference servers. None of these solutions provide the desired development surface for adding application capabilities.
Franca aims to bridge this gap by providing a Java 21 library that handles both dialects within a developer's own service. It turns the server half of the protocol into a standard development surface. Franca supports any inference backend, currently an OpenAI-compatible chat driver, and allows developers to define their tools as ordinary Java beans.
Clients interact with the server using standard OpenAI-compatible endpoints, while Franca handles the server-side implementation details.
Written by urgent.news from Dev.to's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
