Confluent announced Streaming Agents, a new capability in Confluent Cloud for Apache Flink that makes it easy to build and scale AI agents that monitor, reason, and act on real-time data. Streaming Agents removes barriers to enterprise-grade agentic artificial intelligence (AI) by unifying data processing and AI workflows and providing easy, secure connections to every part of a business, including large language models (LLMs) and embedding models, tools, and other systems. It accelerates the adoption of agentic AI, enabling more efficient workflows, faster time to value, and the creation of entirely new business models and opportunities.
“Agentic AI is on every organisation’s roadmap. But most companies are stuck in prototype purgatory, falling behind as others race toward measurable outcomes,” said Shaun Clowes, Chief Product Officer at Confluent. “Even your smartest AI agents are flying blind if they don’t have fresh business context. Streaming Agents simplifies the messy work of integrating the tools and data that create real intelligence, giving organisations a solid foundation to deploy AI agents that drive meaningful change across the business.”
IDC research shows that while organisations ran an average of 23 generative AI proofs of concept between 2023 and 2024, only three reached production. Of those, just 62% met expectations. “While most enterprises are investing in agentic AI, their data architectures can’t support the autonomous decision-making capabilities these systems require,” said Stewart Bond, Vice President of Data Intelligence and Integration Software at IDC. “Organisations should prioritise agentic AI solutions that offer easy, secure integration and leverage real-time data for the essential context needed for intelligent action.”
Build and Scale Real Time AI Agents With Streaming Agents
Streaming Agents brings agentic AI directly into stream processing pipelines to help teams build, deploy, and orchestrate event-driven agents with Apache Kafka® and Apache Flink®. By unifying data processing and AI reasoning, agents gain access to fresh contextual data from real-time sources to quickly adapt and communicate with other agents and systems as conditions change. Streaming Agents are always on and works on a business’s behalf, operating dynamically, processing high-volume data streams, and instantly responding to real-time signals with context-aware reasoning like human operators would. For example, Streaming Agents can do competitive pricing by continuously monitoring prices across ecommerce sites and automatically updating product prices on a retailer’s site to reflect the most competitive offer for customers.
Key features of Streaming Agents include:
- Tool calling for context-aware automation: Tool invocation via Model Context Protocol (MCP) enables agents to select the right external tool, such as a database, software-as-a-service (SaaS), or API, to take meaningful action. Tool calling accounts for what’s happening in the business and what other systems and agents are doing.
- Connections for secure integrations: Securely connect to models, vector databases, and MCP directly using Flink. Connections also protect sensitive credentials, encourage more reusability by sharing connections across multiple tables, models, and functions, and centralise management for large-scale deployments.
- External Tables and Search to boost AI accuracy: Ensure that streaming data is enriched with non-Kafka data sources, such as relational databases and REST APIs, to provide the most current and complete view of data. This improves the accuracy of AI decision-making, vector search, and retrieval-augmented generation (RAG) applications, reduces cost and complexity by using Flink SQL, and leverages the security and networking capabilities of Confluent Cloud.
- Replayability for iteration and safety: Agents can be developed and evaluated using real data without live side effects, enabling dark launches, A/B testing, and faster iteration.
Discussion about this post