Eon Yukon is a managed data engine for teams that need streaming data ready for analytics, live applications, and AI agents without standing up another processing stack. It ingests events, applies SQL transforms, deduplication, and upserts, and keeps Apache Iceberg tables compacted and sorted as data arrives, so query-ready tables are available within seconds. Agents also receive context written in that same pass, including column statistics, change history, and table descriptions exposed over MCP, which is the gap Eon is targeting as models and agents start asking questions that dashboard-era pipelines were not built to answer. Tables land in the customer's own Amazon S3 or Google Cloud Storage and stay readable by Databricks, Snowflake, BigQuery, Athena, Trino, and Spark. Serverless SQL and Spark Connect run next to the data and stop when idle. Configured aggregates can be read by key in under a millisecond through a Redis-compatible API, so applications do not need a separate job to copy state into a cache. Yukon can run on its own or next to Eon Data Protection, the company's cloud backup product, which already makes protected history available as Iceberg tables. In Eon's public benchmark at one million events per second, Yukon used about four times less compute than Apache Spark and about seven times less than Apache Flink at the same freshness. Advertising platform Rise is already running the engine on production traffic in the hundreds of billions of events per day. The launch was posted by CEO and co-founder Ofir Ehrlich, who built Eon with Ron Kimchi and Gonen Stein, the team behind CloudEndure and later AWS disaster recovery and migration services.