Singapore – Couchbase has introduced new capabilities to its AI Data Plane, including native Amazon Bedrock integration with semantic caching, managed agent memory and a managed Model Context Protocol (MCP) server on AWS.
The updates also add vector search to Couchbase Operational Insights. The company said the capabilities are designed to consolidate data services on a single governed platform and reduce the infrastructure complexity and costs associated with running AI agents in production.
The new services target enterprises deploying AI agents across public cloud environments, private data centres and edge devices, including scenarios where connectivity is intermittent or unavailable.
Couchbase cited industry figures indicating that automated requests accounted for 57.5% of web requests in June 2026. It also said AI agents generate 450% more data-layer traffic per task than humans performing equivalent work, while 83% of organisations report needing infrastructure upgrades before they can run production-grade agentic AI.
The company said its distributed architecture is designed to handle concurrent, low-latency operations. Its platform combines key-value storage, SQL, full-text search, vector search and real-time columnar analytics on a single JSON data set, allowing applications to access operational and historical context without relying on separate systems.
Couchbase Mobile also enables applications to run on devices and at the edge, with database and vector search capabilities available locally. The company said its cross-data-centre replication technology keeps regions synchronised and writable.
Mavenir Networks, which provides voice, messaging and data services, has used Couchbase for subscriber profiles and live session state since 2017. The company is expanding its deployment to support its AI initiatives.
“Couchbase has been a key component in achieving high availability in Mavenir’s voice solutions for years,” said Brandon Larson, SVP & General Manager, Cloud, AI & IMS Business Strategy at Mavenir Networks.
“Now we’re creating new AI voice services that require getting reliable context fast and Mavenir is building that context layer with Couchbase because of their proven performance in production,” Larson continued.
Amazon Bedrock integration and semantic caching
The expanded Model Service integrates natively with Amazon Bedrock, providing access to foundation models through serverless, pay-per-token services, alongside vectorisation over private networking.
Couchbase said inference and embedding generation can run alongside data without requiring customers to manage separate GPU instances. The integration also supports semantic caching, which allows applications to reuse results for semantically similar queries rather than calling a model again.
The company said semantic caching has demonstrated the potential to reduce large language model calls by 40% to 60%, depending on the implementation.
Managed agent memory and MCP services
Couchbase has made its Agent Memory and enterprise-supported MCP server available as managed cloud services on AWS, following their initial release as self-managed capabilities in June.
Agent Memory stores information such as user profiles, conversation histories and vectorised facts across sessions, allowing agents to retain context between interactions. The managed MCP server provides a governed route for agents and tools to access Couchbase data without customers having to deploy and operate the server themselves.
Both services are designed to work across different agent development frameworks. Couchbase said they have been validated with LangGraph, CrewAI and LlamaIndex.
The company said Agent Memory can reduce token usage by 33% to 45%, while vector-based context retrieval can produce prompts that are 70% smaller than full-context alternatives. The reductions are intended to lower AI processing costs by limiting the amount of information sent to models.
Vector search added to Operational Insights
Couchbase Operational Insights now supports vector search alongside structured analytical queries on operational data. The combination allows organisations to analyse structured information and identify semantic patterns in unstructured data within a single query, according to the company.
For example, businesses could examine revenue by region while filtering results based on clusters of dissatisfied customers. Couchbase said this can be done without an extract, transform and load (ETL) pipeline or a separate system to synchronise the data.
The platform’s write-back function can also send query results into operational data used by applications and agents, allowing insights to be incorporated into subsequent actions.
Couchbase said customers that consolidate fragmented data services onto its platform have reported total cost of ownership reductions exceeding 50% in more than half of deployments. Early adopters of its AI Data Plane have also reported shortening production timelines from quarters to weeks, the company added.
“Customers building production AI agents at scale need a converged data layer, but most are still running three or four separate systems for memory, vectors, operational data, and analytics, each carrying its own sync logic, security model, and failure mode,” said Barry Morris, Chief Product and Strategy Officer at Couchbase.
“What we’re shipping now collapses that stack. The new capabilities all sit on one governed platform that runs identically in the public cloud, a private data centre, or at the edge. This is helping Couchbase customers reduce costs, accelerate productivity, and realise faster ROI on AI initiatives,” Morris continued.
