Alibaba Cloud, Ververica, Confluent, and LinkedIn Join Forces on the Streaming AI Agents Innovation with Apache Flink®

A landmark collaboration to build scalable, production-grade framework for event-driven streaming agents powered by Apache Flink.

Today at Flink Forward Barcelona 2025, we announced the first release of Apache Flink Agents, a major collaboration between Alibaba Cloud, Ververica, Confluent, and LinkedIn—four influential companies in the Streaming Data area—to jointly develop and contribute a new open-source sub-project from the Apache Flink community designed to bring AI agents into the world of real-time, event-driven systems. This initiative marks a pivotal step toward industrial-scale AI applications that react instantly and autonomously to live data streams.

Why Apache Flink Agents Matters

While AI agents have made rapid progress in interactive applications like chatbots, most still operate outside the high-throughput, low-latency world of real-time data processing. Yet in industrial settings, from e-commerce and finance to IoT and logistics, critical decisions must be made instantly in response to live events: a payment failure, a sensor anomaly, a user click.

These workloads demand more than just intelligence, they require massive scale, millisecond latency, fault tolerance, and stateful coordination, all of which are strengths of Apache Flink. But until now, there’s been no unified framework to bring agentic AI patterns into this proven streaming ecosystem. Apache Flink Agents bridges this gap.

Introducing Apache Flink Agents

Apache Flink Agents, a brand-new sub-project from the Apache Flink community, is an open-source framework for building event-driven streaming agents. Building on Flink's battle-tested streaming engine, Apache Flink Agents inherits distributed, at-scale, fault-tolerant structured data processing and mature state management, and adds first-class abstractions for Agentic AI building blocks and functionalities - large language models (LLMs), prompts, tools memory, dynamic orchestration, observability, and more.

This initiative is the result of a community-based joint effort by developers from Alibaba Cloud, Ververica, Confluent, and LinkedIn, a group of engineers with deep expertise in large-scale stream processing and real-time AI. By combining our experience in production-grade data infrastructure and intelligent systems, we are aligning on a shared vision: bringing Agentic AI into the streaming data ecosystem, where it can operate with scalability, reliability, and real-time responsiveness.

The key features of Apache Flink Agents include:

• Massive Scale and Millisecond Latency: Processes massive-scale event streams in real time, leveraging Flink's distributed processing engine.
• Seamless Data and AI Integration: Agents interact directly with Flink's DataStream and Table APIs for input and output, enabling a smooth integration of structured data processing and semantic AI capabilities within Flink.
• Exactly-Once Action Consistency: Ensures exactly-once consistency for agent actions and their side effects by integrating Flink's checkpointing with an external write-ahead log.
• Familiar Agent Abstractions: Leverages well-known AI agent concepts, making it easy for developers experienced with agent-based systems to quickly adopt and build on Apache Flink Agents without a steep learning curve.
• Multi-Language Supports: Provides native APIs in both Python and Java, enabling seamless integration into diverse development environments and allowing teams to use their preferred programming language.
• Rich Ecosystem: Natively integrates mainstream LLMs, vector stores from diverse providers, and tools or prompts hosted on MCP servers into your agents, while enabling customizable extensions.
• Observability: Adopts an event-centric orchestration approach, where all agent actions are connected and controlled by events, enabling observation and understanding of agent behavior through the event log.

Looking ahead

This initial version provides core agent abstractions, integrates Flink's DataStream and Table APIs, supports Kafka-based action consistency, integrates with selected LLMs and vector stores, includes MCP support, and offers observability through event logs.

This milestone marks the beginning of a powerful new framework for building scalable, event-driven AI agents on top of Apache Flink. For more information, visit the Apache Flink GitHub repository.

Detailed timelines, design discussions, and community input can be found in the GitHub Discussions section your go-to place to follow and contribute to the project’s evolution.

If you're passionate about building intelligent, autonomous systems that react in real time to streaming events, Flink Agents is a project worth watching and joining. We warmly invite developers, contributors, and AI enthusiasts to get involved and help shape the future of event-driven AI with Apache Flink.

Learn more

• Read our in-depth blog: Flink Agents: An Event-Driven AI Agent Framework Based on Apache Flink
• Watch the presentation from Flink Forward Asia Singapore 2025: Flink Agents – The Agentic AI Framework based on Apache Flink
• Explore the original proposal: FLIP-531: Initiate Flink Agents as a new Sub-Project

The future of AI isn’t just smarter models, it’s smarter systems that act continuously, reliably, and at scale.
With Apache Flink Agents, we’re building that future together.