Couchbase launches managed agent memory and MCP server
Couchbase announced on October 20, 2026, the general availability of two managed services for the AI agent economy: Managed Agent Memory and a Managed MCP Server, both available as fully managed offerings on AWS within Couchbase Capella, the company's cloud data platform Couchbase.
Managed Agent Memory for Stateful Agents
The Managed Agent Memory layer provides a durable, low-latency persistence tier for agent profiles, conversation history, and vectorized facts that persist across sessions Couchbase. This capability addresses a core operational requirement for production agents: maintaining context and learned behavior without rebuilding state on every invocation. Both capabilities were released as self-managed options in June 2026 and migrated to fully managed cloud services by October.
Couchbase validated the services against three major agent frameworks: LangGraph, CrewAI, and LlamaIndex, indicating framework-agnostic support Couchbase.
Governed Data Access via Managed MCP Server
The second release—a Managed MCP Server—implements the Model Context Protocol as a governed access layer for agents to retrieve and interact with Couchbase data. The managed server runs as a single-tenant instance within a customer's Capella environment, eliminating the need for customers to provision and operate separate server infrastructure Couchbase. The server includes more than 20 tools for schema introspection, query execution, and cluster health monitoring, and operates in read-only mode by default Couchbase.
MCP, an open standard developed by Anthropic, allows language models and agents to access external data sources and tools through a standardized protocol. Couchbase's managed implementation reduces operational overhead for enterprises deploying agents at scale.
Semantic Caching and Vector Search Integration
The October release also integrates native Amazon Bedrock connectivity with semantic caching, allowing agents to avoid redundant API calls and reduce inference costs. Vector search capabilities are now available in Couchbase Operational Insights, Couchbase's analytics and observability layer, enabling semantic queries across agent-generated embeddings and unstructured data Couchbase.
These capabilities address two persistent pain points in production agent deployment: persistent state management and cost control. By bundling memory, data access governance, and semantic search in a single managed service, Couchbase positions its platform as infrastructure for agents that operate autonomously across sessions and require reliable, cost-controlled access to enterprise data.
