Confluent Debuts Streaming Agents for Real-Time Agentic AI

Data streaming company Confluent has introduced Streaming Agents, a new capability on Confluent Cloud for Apache Flink that enables enterprises to build and scale real-time AI agents. Now in open preview, the feature unifies stream processing and AI reasoning so that agents can monitor, decide and act on live business events with contextual awareness.

Unlike traditional AI workflows that depend on static snapshots, Streaming Agents run as always-on event-driven microservices. Embedded directly into Flink pipelines, they continuously ingest, reason, and respond to high-volume data streams. Every action is backed by an immutable event log, giving teams replayability for testing, debugging and audit trails.

The capability includes tool calling through Model Context Protocol (MCP), allowing agents to invoke the right external tools, databases, SaaS applications or APIs, based on real-time business context. 

Connections provide secure integration with models, vector databases and third-party systems, with centralised credential management to support enterprise-scale deployments.

Streaming Agents also enrich streaming data with non-Kafka sources such as relational databases and REST APIs, enhancing accuracy for retrieval-augmented generation (RAG) and decision-making. 

As per the blog post, real-time embeddings convert unstructured enterprise data into vector form for semantic search, helping mitigate hallucinations and maintain the freshness of context. Built-in machine learning functions, including forecasting and anomaly detection, allow teams to simplify data science tasks directly within Flink SQL.

“Even your smartest AI agents are flying blind if they don’t have fresh business context,” Shaun Clowes, chief product officer at Confluent, said. “Streaming Agents simplifies the messy work of integrating the tools and data that create real intelligence.”

Confluent highlighted enterprise applications such as competitive pricing, anomaly investigation and real-time product personalisation, positioning Streaming Agents as a path to production-ready, event-driven multi-agent systems.