With Protocol Buffers, a developer writes a .proto file describing message fields and their numbers, and a compiler generates serialization code for many languages. On the wire each field is encoded as a number and a value, with no field names, so messages are small and fast to parse, and the schema guarantees that producer and consumer agree on types.
In IoT it appears wherever data volume or structure matters: Sparkplug uses it for MQTT payloads, gRPC uses it for service APIs between edge and cloud, and embedded implementations such as nanopb bring it to microcontrollers. The trade-off is the schema itself. Every party must have the right definition, adding fields requires discipline about numbering, and payloads cannot be read without it, so debugging is harder than with JSON. CBOR offers compactness without a mandatory schema.
Part of the IoT Glossary. Spotted an error or a missing term? Tell us.