What is Change data capture (CDC)?

Change data capture, CDC, is the practice of turning the changes happening in a source database into a continuous stream of events that downstream systems can consume. Rather than periodically re-querying a table to find what changed, which is slow, misses fast updates, and loads the database, CDC reads the database's own transaction log (the write-ahead log in Postgres, the binlog in MySQL) and emits an event for each row inserted, updated, or deleted, in order, as it commits. That log-based approach is low-overhead and lossless, which is why it has become the standard way to keep systems in sync: replicating into a data warehouse, invalidating caches, populating a search index, feeding analytics, or driving event-driven workflows. For AI systems CDC is increasingly relevant for keeping retrieval fresh, when a source row changes, a CDC pipeline can trigger re-embedding so a vector store or RAG index reflects current data instead of a stale snapshot. Tools like Debezium popularized log-based CDC, and managed databases expose it as a feature; the main design concerns are ordering guarantees, handling schema changes, and the eventual-consistency lag between a write and its propagation downstream.