Unleash Real-Time Agentic AI With Streaming Agents on Confluent Cloud and Couchbase

We’re thrilled to be partnered with Confluent today as they announce the new features for Streaming Agents on Confluent Cloud and a new Real-Time Context Engine. Unifying data processing and agentic AI workflows, Streaming Agents enable developers to build, deploy, and orchestrate event-driven agents using fully-managed Apache Flink® and Apache Kafka® on a unified platform. Today’s new capabilities take this further by helping teams build trustworthy agents faster and more easily, gain enhanced observability, and improve AI decision-making with real-time context.

As a Confluent Partner, we see firsthand why the introduction of Streaming Agents was so critical. At its core, every AI problem is a data problem. When data is stale, incomplete, or inaccessible, even the most sophisticated agents and large language models (LLMs) can’t deliver reliable results.

This is exactly what the market demanded: a solution for building scalable multi-agent systems that are event-driven, replayable, and grounded in fresh, contextualized data. Developers needed a single platform that didn’t just provide isolated tools, but one that enabled them to easily source data, reliably move from prototype to production, and gain the necessary observability to debug, evaluate, and iterate on what’s actually happening inside their agents.

Streaming agents power intelligent, context-aware automation

Embedded in data streams with access to the latest, most complete and accurate view of operational events, Streaming Agents effectively act as the “eyes and ears” of the business. They’re designed to handle high-volume real-time data and evolving context, making them ideal for enterprise use cases where fresh information, accuracy, and observability are critical. By continuously monitoring data streams and using context from diverse sources, Streaming Agents can make intelligent, informed decisions and automate actions that drive better outcomes.

High-value use cases include:

- - Real-time fraud prevention – Continuously ingest and process transaction data, detect anomalies, and automatically block suspicious activity.
  - Intelligent supply chain optimization – Track inventory, shipments, and demand signals in real time, automatically reordering stock, rerouting shipments, or adjusting production schedules based on live conditions.
  - Dynamic customer support – Pull live context from CRM systems, chat interactions, and knowledge bases to deliver in-the-moment personalized and accurate responses.

Let’s explore this last customer support use case in more detail. Imagine a global enterprise streaming customer chat events, CRM updates, product inventory signals, and service agent feedback into Confluent Cloud. A Streaming Agent built on Flink:

- - Consumes the live chat event
  - Enriches it with context (customer history, latest purchase, open tickets)
  - Embeds the chat via an embedding model
  - Performs a millisecond vector search lookup in Couchbase 8.0 to find semantically relevant past conversations, knowledge base articles, and support actions
  - Invokes a tool via MCP in real time (e.g., ticket update API, service scheduling tool)
  - Generates a response via an LLM with RAG support
  - Feeds the result back into Couchbase (update of conversation state) and Kafka topic for audit/analytics

The result: a customer support agent that is context-aware, real-time, semantic+vector powered, automated, and fully observable. The underlying data stack on Couchbase ensures the freshest content and semantic retrieval, the Confluent streaming engine ensures event-driven flow, tool orchestration, and production-grade scale.

What’s new in the Q4’25 release

With Streaming Agents, every engineer can use familiar Flink APIs to build secure and trustworthy agents, with native support for Model Inference, Tool Calling with MCP, Embeddings for RAG, Built-in ML Functions, External Tables and Search, and Connections. Confluent is continuing to expand on these capabilities and deliver more streamlined developer experiences.

- - Agent definition – Quickly build agents in just a few lines of code and unlock more sophisticated tasks with better outcomes by iteratively evaluating and adapting tool calling.
  - Observability and debugging – Gain visibility into all agent actions, easily diagnose issues to accelerate resolution, and reliably recover from failure.
  - Real-time context engine – Using MCP (Model Context Protocol), serve fresh context to Streaming Agents as well as any other AI agent and application to improve decision-making and the quality of outputs.

Synergy with Couchbase 8.0: A unified data platform for Agentic AI

While the streaming-agent framework provides the orchestration and logic for agentic AI, the underlying data platform is just as critical. That’s where Couchbase steps in, and the timing could not be better with the recent launch of Couchbase 8.0. Here’s what you should know:

- - Hyperscale vector indexing: Couchbase 8.0 introduces supporting billion-scale vector search workloads with millisecond latency and tunable recall accuracy. Independent benchmarking showed more than 19,000 queries per second at 28 ms latency (66% recall) and strong results at higher recall settings.
  - Unified workload support: Vector search is not an add-on—it’s part of the same platform that handles key-value access, JSON documents, distributed caching, search (vector, text, GEO.), analytics, and mobile sync for offline-first apps. That means operational, AI/agentic, and analytical workloads coexist without stitching multiple data silos.
  - Real-time context, freshness, and trust: Agentic AI depends on timely, accurate context. If the vector retrieval layer is stale, disconnected or high-latency, the downstream agents suffer. Couchbase 8.0 strengthens the ability to provide fresh embeddings, real-time document updates, and live index refreshes, which are core to the streaming agent pattern.

Streaming agents + Couchbase = Real-time Agentic AI at scale

Here’s how the partnership plays out, and why we believe it provides a compelling foundation for next-gen enterprise agentic systems:

1. 1. Real-time ingestion & streaming context: With Confluent Cloud running Kafka + Flink + Streaming Agents, operational events are captured, processed, enriched and transformed in real time.
  2. Streaming Agents embed AI workflows: Developers use Flink APIs to embed ML functions, call tools, invoke LLMs or other models, vectorize unstructured content, join streaming context with external tables, and orchestrate workflows.
  3. Vector search feed from Couchbase: The latest data, embeddings, document updates and context live in Couchbase. Streaming Agents can link to the Couchbase vector index to supply semantic + contextual retrieval to agents, thus powering RAG, contextual chatbots, real-time decision logic, anomaly investigation, etc.
  4. Closed-loop, adaptive agentic systems: The streaming pipeline can feed back agent outcomes and updates into Couchbase. Over time, agents learn, adjust, and the context store updates. The unified platform supports production-scale, operational applications, not just one-off ML pipelines.