Couchbase on Tuesday announced the AI Data Plane, a platform that combines persistent agent memory, real‑time context retrieval and an enterprise‑managed MCP server into a single operational offering. The company positions the package as a way to replace the fragmented stacks many enterprises currently use.
What the AI Data Plane includes
- Agent memory: a unified persistence layer for conversational context, structured operational data and vector embeddings. Couchbase says it differentiates from standalone memory services by built‑in guardrails: token limits per session, time‑to‑live constraints on stored memories and metering controls that cap compute consumption per agent session.
- Enterprise MCP server: a self‑managed, enterprise‑supported server for standardized model‑context protocol integration shipped as part of the platform rather than as a separate service.
- Agent catalog: a function‑level catalog of discoverable agent tooling built by Couchbase. Gopi Duddi, Chief Technology Officer at Couchbase, distinguished this from metadata catalogs like Databricks Unity or AWS Glue, describing it as closer to a glorified MCP that surfaces agent functions as callable tools within the platform.
Memory‑first architecture and the disconnected edge
Couchbase points to its origins in caching and high‑transaction databases as the architectural reason it is well suited for agent memory. Duddi noted that they "were a cache before we became a database," and said writing to memory is ten times faster than writing to disk — a performance advantage he argues separates Couchbase from NoSQL databases that layer memory workloads on top of disk‑based storage.
That architecture extends to the edge via Couchbase Lite, the on‑device runtime. Couchbase Lite can run SQL, full‑text search and vector search locally without a network connection, and uses a proprietary sync mechanism to replicate bidirectionally back to cloud or between edge nodes when connectivity returns. Target environments include retail floor operations, field service, industrial deployments and regulated settings where agent data must remain on the device.
Duddi cited hotel reservations as an early example: multiple agents serving customers concurrently, each pulling local context and running on‑device vector search, with shared session memory synchronizing centrally. The practical benefit is token efficiency: instead of each agent independently retrieving and reprocessing the same data, the platform caches shared context so concurrent sessions draw on it without consuming tokens repeatedly.
In production: Agora's experience
Agora, a platform that helps developers embed real‑time voice, video and conversational AI into enterprise applications, has run Couchbase in production since February 2024. The initial use case was the Signaling product for channel setup and state synchronization for live calls. Expanding into conversational AI agents introduced stricter requirements: a memory‑first architecture, full JSON support for storage and query, cross‑datacenter replication for high availability and enterprise‑grade vendor support.
Patrick Ferriter, SVP of Product at Agora, told VentureBeat that "Couchbase was the best fit based on these criteria." Agora is now extending that relationship to support context retrieval for conversational AI agents. Ferriter said, "This will simplify the architecture and deliver enterprise grade RAG with predictable lower latency required for conversational AI use cases."
Ferriter also noted there is no single right approach for data teams choosing context solutions: platform selection depends on organizational preferences and goals, including timing and whether a team prefers an enterprise‑grade commercial product or an open‑source system requiring optimization and maintenance.
Competitive landscape in 2025
The context layer market became crowded in 2025. Oracle added a "memory core" to its database in March, Redis announced a context layer in May, and vector‑native database vendor Pinecone has also moved into the space. IDC research director Devin Pratt said Couchbase is following the trend rather than setting it, but its strength lies in reach: running the same platform across cloud, on‑prem and mobile/edge environments — how enterprises actually operate. The key test will be scaling against larger incumbents.
Pratt advised teams to match tools to workloads: consolidate where it makes sense, use specialized engines (for example, a graph database for relationship‑heavy reasoning), and let governance drive decisions rather than treating memory merely as plumbing.
Conclusion
Couchbase's AI Data Plane offers an integrated approach combining persistent agent memory, real‑time retrieval and a managed MCP server that operates uniformly across cloud, on‑premises and offline edge. Early production use at Agora and recent moves from other vendors frame Couchbase as a contender that emphasizes memory‑first architecture and deployment reach, while its broader market success will depend on comparative scalability and enterprise adoption.
