Couchbase outlined a strategy to converge the agent data stack onto its operational platform. It now offers Agent Memory as a managed service on AWS. The service handles conversation history, user profiles and semantically searchable facts. Couchbase also launched a managed MCP server on AWS for standard agent access to live operational data and tools. Vector search and write-back were added to Operational Insights to connect analysis with operational action. A native Amazon Bedrock integration provides pay-per-token model access with semantic caching. Couchbase estimates cost reductions from caching, memory and vector queries. Forrester advises enterprises to model operational state and decision history alongside business meaning. It also recommends serving context through governed, open interfaces. Enterprises should capture write-backs with their provenance. They must assign owners to keep context synchronized and current.
What changed
Agent data stacks were built from separate services requiring manual integration and data synchronization.
What it unlocks
Running agents that combine operational and analytical data, persistent memory, semantic retrieval, and tool access in a single reasoning loop on one platform.
- 40% to 60% fewer model calls from semantic caching
- 33% to 45% lower token use from Agent Memory
- 70% smaller prompts from vector query responses
What you need to act on it
- AWS deployment for managed Agent Memory
- MCP-compatible agents for tool access
- Amazon Bedrock account for model access
Sources