The competitive advantage in enterprise AI is shifting to context: which platform can give an agent the right memory, the right retrieval, and the right data at the time of making a decision.
Couchbase on Tuesday announced its AI Data Plane, which combines persistent agent memory, real-time context recovery, and an enterprise-managed MCP server into a single operating platform.
Couchbase’s roots are in caching and high transaction databases – an architecture that the company says makes it more suitable for agent memory than providers who arrived at the problem through search or analysis. AI Data Plane runs identically in cloud, on-premises, and disconnected edge environments, extending agent memory and local vector search to devices without a network connection.
"How do you ensure that the intelligence you get from these models is what the databases specialize in?" Gopi Duddi, CTO of Couchbase, told VentureBeat. "How can you get that value out of storage systems, which will still be databases?"
What AI Data Plane offers
AI Data Plane includes three components designed to replace the fragmented stacks that most companies currently run.
Agent memory: A unified persistence layer for conversational context, structured operational data, and vector embeddings. Couchbase says security barriers are what distinguish it from standalone memory services: per-session token restrictions, lifetime limits on stored memories, and metering controls that limit compute consumption per agent session.
Enterprise MCP Server: A self-managed server with enterprise support for the integration of standardized model context protocols, shipping as part of the platform rather than requiring a separate service.
Agent catalog: A feature-level catalog of recognizable agent tools created by Couchbase. Duddi distinguished it from metadata catalogs like Databricks Unity or AWS Glue and described it, in his words, as closer to a glorified MCP that displays agent functions as callable tools within the platform.
Memory-centric architecture brings agent context to the disconnected edge
Couchbase’s lineage and core architectural foundation is what Duddi says gives it an advantage when it comes to context.
"We were a cache before we became a database." Duddi said.
Writing to memory is 10 times faster than writing to disk, Duddi said; a speed advantage that he says separates Couchbase from NoSQL databases that put in-memory workloads on top of disk-based storage.
Couchbase is not the only data technology that has its roots in a caching layer. Redis is similarly rooted in the cache and also recently announced an AI agent context layer. Duddi argued that Couchbase differs in that it maintains an ACID (Atomicity, Consistency, Isolation, and Durability) compliant database that is important for transactional workloads. Couchbase also has a long history of multiple deployment modalities.
That architecture extends to the edge through Couchbase Lite, the platform’s on-device runtime. Runs SQL, full-text search, and vector search locally without a network connection, using a proprietary synchronization mechanism to replicate bidirectionally to the cloud or between edge nodes when connectivity returns. Target environments are retail operations, field services, industrial deployments, and regulated environments where agent data cannot leave the device.
Duddi cited hotel reservations as an early example: multiple agents serve customers simultaneously, each extracting local context and running a vector search on the device, with a shared session memory that is synchronized centrally. The practical benefit is symbolic efficiency. Instead of each agent retrieving and processing the same data independently, the platform caches the shared context for concurrent sessions to use without repeatedly burning tokens.
Agora’s vision from production
Agora, a platform that helps developers integrate real-time voice, video, and conversational AI into enterprise applications, has been running Couchbase in production since February 2024.
The initial use case was their signaling product, which manages channel configuration and status synchronization for live calls. The expansion to conversational AI agents brought more stringent requirements: memory-first architecture, full JSON support for storage and queries, cross-datacenter replication for high availability, and enterprise-grade vendor support.
"Couchbase was the best option based on these criteria," Patrick Ferriter, senior vice president of product at Agora, told VentureBeat.
Agora is now expanding that relationship to support context retrieval for conversational AI agents.
"This will simplify the architecture and deliver enterprise-grade RAG with lower, more predictable latency needed for conversational AI use cases." Ferriter said.
For data professionals trying to find the best approach for context, there is no single answer. When it came to platform selection, Ferriter was direct.
"It depends on the preferences and objectives of the organization, including the timing," Ferriter said. "If they want something enterprise grade and optimal for production and immediate scale instead of having to optimize and maintain an open source solution with community support. We wanted the former and that’s why we considered an expanded partnership with Couchbase."
Competitive context: follow the right trend
The context layer has become a crowded space in 2025.
Oracle put a memory core in its database in March providing a layer of context. Redis added a context layer in May just like the vector native database provider Pineapple.
"Couchbase follows this trend, it doesn’t set it, but it is the right one to follow." Devin Pratt, research director of AI, automation, data and analytics at IDC, told VentureBeat. "Their real advantage is reach, running the same platform from cloud to edge to mobile, which is how businesses really operate. The test now is to climb against bigger names."
For teams navigating the supplier landscape, Pratt’s approach is straightforward. "Match the tool to the workload. Consolidate where it makes sense, use a specialized engine like a graph database where relationship-based reasoning deserves it, and let governance drive the call instead of treating memory like a pipeline." Pratt said.
