Couchbase AI Data Plane

Couchbase AI Data Plane is a unified operational data layer for production AI agents that combines persistent agent memory, vector search, semantic caching, an agent tool catalog and a supported MCP server in a single platform. It is built for CTOs and platform teams tired of stitching together separate vector, cache and document stores for every agent they ship.

Couchbase announced general availability of the AI Data Plane on June 30, 2026, repositioning its database as the operational foundation for agentic enterprises. The product — formerly marketed as Couchbase AI Services — unifies four named components: Agent Memory, which provides persistent context that survives sessions and restarts and supports short-term, long-term, semantic and conversational memory types; an enterprise-supported, self-managed MCP server for standardized Model Context Protocol data access; Agent Catalog, which manages prompts, tools and end-to-end traces so agent behavior is discoverable and governable via SQL++; and AI Functions, SQL++ and LLM utilities for tasks such as sentiment analysis. It runs on Couchbase's distributed multimodel architecture — JSON documents, key-value and SQL-for-JSON queries — and adds billion-scale vector search, semantic caching to cut redundant LLM calls, and automated data ingestion and vectorization. Couchbase's pitch is cost and latency: Agent Memory reduces token spend by reusing persistent cumulative context instead of resending whole conversations each turn, while co-located data and caching cut redundant inference calls at sub-millisecond latency. The release consolidates previous Couchbase deployment models into one architecture spanning Capella DBaaS, self-managed on-premises and multicloud, hybrid, edge and air-gapped environments, and is framework-agnostic — validated with LangGraph, CrewAI and LlamaIndex. It ships alongside Enterprise Analytics 2.2, which adds Apache Iceberg lakehouse federation, JWT authentication, change data capture for Oracle and SQL Server, an index advisor and SQL++ UPDATE support; a Trino adapter enabling SQL access via AWS Athena, Amazon EMR, Google Dataproc and Starburst is expected in Q3 2026. Chief Product and Strategy Officer Barry Morris framed the launch bluntly: 'The database layer is where agentic AI either scales or stalls, and most of the industry is still treating agent memory as an afterthought.'

Category

Infrastructure & Cloud

Pricing

Freemium, Usage-based, Contact for pricing

Target Market

CTOs, Enterprise Developers, Data Engineers, Platform Engineering Teams, AI Engineers

• ✓
  Agent Memory

  Persistent agent memory across sessions and restarts supporting short-term, long-term, semantic and conversational memory types.

• ✓
  Enterprise MCP server

  An enterprise-supported, self-managed Model Context Protocol server providing standardized data access for external AI agents.

• ✓
  Agent Catalog

  Manages prompts, tools and end-to-end traces, giving governance and visibility into agent behavior queryable through SQL++.

• ✓
  Billion-scale vector search with semantic caching

  Vector search over structured and unstructured data at billion scale, with semantic caching to reduce redundant LLM calls.

• ✓
  AI Functions

  SQL++ and LLM utilities that run tasks such as sentiment analysis directly against operational data.

• ✓
  Cloud-to-edge deployment

  One architecture across Capella DBaaS, self-managed on-premises and multicloud, hybrid, edge and air-gapped environments.

• •
  Consolidating fragmented agent data infrastructure

  Replaces the separate vector, caching and document stores teams previously stitched together for every agent, which Couchbase cites as the biggest drag on production timelines.

• •
  Reducing agentic AI token and inference cost

  Agent Memory reuses persistent cumulative context instead of resending conversations each turn, while co-located data and caching cut redundant inference calls.

• •
  Operational analytics on agent and enterprise data

  Enterprise Analytics 2.2 federates Apache Iceberg lakehouses and adds change data capture from Oracle and SQL Server for real-time operational querying.

A Couchbase Capella free tier and an Agent Memory trial are available; enterprise pricing for the AI Data Plane was not disclosed at launch.

This page was written from 3 sources, 2 on domains other than couchbase.com.

1. 1.couchbase.com — ai servicesvendor
2. 2.blocksandfiles.com — 5264312
3. 3.sdxcentral.com — couchbase launches ai data plane to unify enterprise agent i