Buf wanted a Go client for TypeSafe’s Jev and there was not one, so it wrote a code generator. In Asking Jev questions with Protobuf, Kevin McDonald introduces protoc-gen-jev, a protoc plugin that “defines Jev questions in Protobuf and generates typed clients and response types for Go, TypeScript, and Python.” The request message carries the state, the response message carries the questions as field options, and Jev’s three primitives map onto ordinary types: a choice becomes an enum, a score becomes an integer on a rubric, and a noul, a yes-or-no probability, becomes a boolean once it clears a threshold the schema sets, 0.7 in the example.
The figures are Buf’s own test run, and the post is careful about them. Against Cloudflare’s Clef-omni on Workers AI, a checkout outage ticket came back technical, frustration 1, urgent, with an urgency probability of 0.92; an angry dark-mode complaint came back technical, frustration 3, not urgent, at 0.07. Two tickets are a demonstration, not an evaluation, but the pair makes the design point well: frustration and urgency are separate questions and the schema keeps them separate. Buf is just as plain about status: “protoc-gen-jev is an experiment, not a supported Buf product,” it runs locally only, and there is no BSR remote plugin or generated SDK yet.
That disclaimer is what makes the Buf provider page the right place to read it. The catalog record describes Buf’s own public API as Protobuf rather than OpenAPI, 33 services and 85 RPCs at the root of buf.build, exposed to agents through a remote MCP server. The post uses the Buf CLI to run buf generate and pulls its option definitions from the Buf Schema Registry (BSR) as a dependency. Shipping the plugin as a remote plugin on the registry is the step Buf says it has not taken, and it is the step that would turn an experiment into something the registry distributes like everything else it hosts.
The Kin Score is 65.3, strong band, built on operational transparency at 92.1, developer ergonomics at 85.7, and access clarity at 85.5, with contract quality at 39.0 and contract governance at 18.2. Agent Readiness is 49.2, agent-ready, with the MCP server, agent skills, error semantics, idempotency, delegated identity, and dynamic client registration lit, and the well-known catalog, OpenAPI examples, and dry-run mode dark. A company whose argument is that a schema should carry the meaning of a call, all the way down to the question an AI model is asked, scores lowest on the facets that measure the schema itself. Part of that is our rubric reading OpenAPI more easily than Protobuf. The governance gap is still worth closing on Buf’s own terms.