COUCHBASE RELEASE
Smarter Agents and Insights, Right Where Your Data Lives
Agent Memory and MCP Server join the managed AI Data Plane services in Couchbase Capella, with serverless models and private embeddings. And Couchbase Analytics is now Operational Insights.
20+
ready-made MCP Server tools, read-only by default
94%
recall on LongMemEval-M single session
1
SQL query to group by semantic meaning and aggregate
The Couchbase you know gets even better for agents
Agent Memory and MCP Server now run fully managed in Couchbase Capella, alongside Agent Catalog and the rest of the AI Data Plane. And Couchbase Analytics has a new name, Operational Insights, with powerful new ways to search and group data by semantic meaning in SQL.
Fully managed
Couchbase installs, patches, and upgrades managed Agent Memory and MCP Server for you.
Same APIs, two deployments
Agent Memory and MCP Server use the same model and APIs whether Capella runs them or you do.
Data stays secure in your perimeter
Embeddings travel over private networking to Amazon Bedrock, and your operational data stays in Couchbase.
Analytics, renamed and upgraded
Operational Insights now supports vector search, semantic clustering, and Python UDFs (self-managed).
New AI Data Plane capabilities, built for production
Streamline deployment of AI agents from pilot to production with persistent Agent Memory, Apache Iceberg lakehouse federation, and Rust support. And for mobile, Bluetooth peer-to-peer sync and React Native support.
Managed Agent Memory
Durable agent memory as a Capella service, with TTL-based decay, conflict resolution, and OIDC/OAuth2 built in.
Managed MCP Server
A remote, single-tenant MCP server in your Capella environment, with ready-made tools and read-only defaults.
Serverless frontier models
Claude, Llama 4, Nova, Mistral, and Cohere from Amazon Bedrock. Pay per token, with no GPUs to size or reserve.
Private Bedrock vectorization
Generate embeddings with your own Amazon Bedrock account over AWS PrivateLink, off the public internet.
Couchbase Analytics is now Operational Insights
Same foundation, new name. Capella Analytics is now Capella Operational Insights, and Enterprise Analytics is now Enterprise Operational Insights. The latest version adds vector search, semantic clustering and Python UDFs (arriving first for self-managed deployments).
Vector search in SQL
Self-managed deployments let you search by meaning in plain SQL. Vectors and a VTREE index keeps approximate search fast at scale. No separate vector database.
Semantic clustering
CLUSTER BY groups by meaning, then counts and aggregates each group in the same query. Find themes in reviews, tickets, or claims without manual tagging (self-managed).
Python UDFs
Write custom functions in Python and call them from SQL on self-managed deployments. Run your own ML models where the data lives, with no export to a separate service.
Faster, leaner queries
Query plan caching skips planning for repeat queries, hash-based aggregation counts DISTINCT values in one pass, and sampled optimizer stats avoid full-table scans.
Multi-statement queries
Send several SQL++ statements in one request, so multi-step analysis needs less application code.
One product, two deployments
Capella Operational Insights is fully managed in the cloud. Enterprise Operational Insights runs on the same foundation on premises or in your own cloud.
One platform for data, memory, and models
Your operational record, agent memory, governed access, models, and insights run on one Couchbase foundation, with SQL++ across operational and analytical data. That means fewer systems to keep in sync and one place to set access rules.
Context that carries over
Agents keep durable facts, decaying context, and semantic recall in one JSON store. Users stop repeating themselves, and your team stops syncing separate memory systems.
Access control in one place
The Couchbase MCP Server gives every agent and tool one open-standard front door to governed Couchbase data, instead of a custom integration per agent where permissions can drift.
Lower cost to run AI
Right-sized context trims token spend on every call, and per-token model pricing means no idle GPUs to pay for.
Insight your agents can use
Write Operational Insights results back to Couchbase, where apps and agents act on them on the next read through the AI Data Plane.
What customers are saying
-
“Couchbase was superior, especially for large workloads.”
Claus Moldt, CIO, FICO<1ms
tempi di risposta
24×365
tempo di attività dell'applicazione
Two businesswomen working on laptop while sitting on couch during coffee break in modern workspace. -
“Real-time communications data and high concurrency query greatly improve the performance and stability for the AI Assistant application.”
Andy Qiu, CEO, Jinmu8K
operazioni al secondo
10ms
response time
-
“As we’ve scaled into agentic workloads, this was a natural extension to our partnership.”
Patrick Ferriter, SVP of Product, Agora<.5
sec target per agent transaction
Volete saperne di più?
Blog: What’s new in the Couchbase October 2026 release
A walk-through of everything that becomes generally available on October 19, from managed AI services to new models and private embeddings.
You’ll learn how:
- Managed Agent Memory and MCP Server run on Couchbase Capella, and what stays the same if you self-manage
- Serverless frontier models from Amazon Bedrock work, including per-token billing in Capella credits
- Private vectorization keeps embedding traffic off the public internet
Video: Introducing Couchbase Operational Insights
Your AI agents are only as good as the data they’re working from. In this video, you’ll see how Operational Insights finds answers in your live data, so agents act on what’s happening now, not on yesterday’s copy.
You’ll see how Operational Insights :
- Asks questions of your JSON data as it is, without flattening it or building a pipeline first
- Spots patterns, not just totals, by grouping records by meaning alongside the usual counts and sums
- Sends answers straight back to your apps and agents, so they always have the latest information
