Couchbase Inc. has expanded its AI Data Plane with new features, including native Amazon Bedrock integration, semantic caching, and an Agent Memory and Model Context Protocol server on AWS. The company also added vector search to its analytics engine, now branded as Operational Insights.
According to Couchbase, semantic caching reduces large language model calls by 40% to 60%. The company claims its Agent Memory reduces token usage by 33% to 45% compared to other benchmarks, while vector-based context retrieval produces prompts that are approximately 70% smaller than full-context methods.
The updates address rising infrastructure challenges as AI agents generate 450% more traffic per task than human users, according to Cisco Systems data. Couchbase aims to mitigate the need for companies to downgrade to cheaper, less capable models by reducing token waste and optimizing data retrieval.
Couchbase argues that consolidating data into a single system reduces latency and security risks compared to using multiple disparate tools. The company reports that its technology is currently deployed at organizations including LinkedIn, Amadeus, and Mavenir.
