Avro

Apache Avro is a data serialization system that pairs a compact binary encoding with an explicit, JSON-defined schema. Created within the Apache Hadoop project in 2009, it was designed for high-volume data pipelines where both size on the wire and the ability to evolve data structures over time matter. Every Avro payload is written against a schema, which is what makes it durable across changing systems.

• Schema-first encoding - Data is always serialized against a schema defined in JSON, so structure is never ambiguous.
• Compact binary format - Field names are not repeated per record, producing much smaller payloads than verbose text formats.
• Schema evolution - Reader and writer schemas can differ, letting producers and consumers add or change fields without breaking compatibility.
• Big-data heritage - First-class support across Hadoop, Kafka, and the broader streaming ecosystem, often paired with a schema registry.

In real API and event-driven operations, Avro most often shows up as the payload format behind message streams and schema registries, where the contract between producer and consumer must stay stable as data models change. It sits in the same design space as other serialization and schema approaches teams weigh when moving structured data at scale, and its explicit-schema discipline is exactly the kind of machine-readable contract that agent-driven systems depend on to understand data without guessing.