Confluent adds A2A support to fuel multi-agent AI networks
Including the open protocol enables users to build an orchestrated network of collaborative agents and could help the vendor distinguish itself from its closest competitors.
After unveiling capabilities last October that enable customers to develop AI agents fed by streaming data, Confluent is taking the next step and now helping users orchestrate multi-agent systems.
With the introduction of support for the Agent2Agent (A2A) protocol in open preview, the vendor's users can connect agents across the organization to transform them from applications capable of performing specialized tasks into a network working together make businesses more intelligent and efficient.
In addition, Confluent unveiled Multivariate Anomaly Detection, a feature now in the early access stage that autonomously examines a series of metrics to discover unusual patterns in streaming data pipelines to surface and prevent issues before they affect application performance.
Both A2A support and Multivariate Anomaly Detection are part of Confluent Intelligence, a managed service for building and deploying AI agents, and are designed to make it easier for users to build context-aware AI systems that evolve along with changes in their source data.
Given their aim of enabling Confluent customers to create cutting-edge AI networks, the new features are significant additions, according to Stephen Catanzano, an analyst at Omdia, a division of Informa TechTarget.
"The introduction of A2A support and Multivariate Anomaly Detection … create a more intelligent and adaptive system for real-time data analysis," he said. "Together, these features empower businesses to act on live signals, automate decisions and prevent issues before they escalate, making them highly valuable for enterprise-scale operations."
Based in Mountain View, Calif., Confluent is a streaming data specialist providing a platform built on the open source Apache Kafka technology. Currently, the vendor is under agreement to be acquired by IBM for $11 billion in a deal unveiled in December 2025 but not yet closed.
Confluent's impetus for adding the A2A came from a combination of customer feedback and market observation while the motivation for developing Multivariate Anomaly Detection was based on interactions with users, according to Sean Falconer, the vendor's head of AI.
"We launched single-variable anomaly detection last year … but some customers ran into scenarios where looking at one signal at a time just wasn't enough," he said. In addition, he noted that "A2A is quickly becoming a standard across agent frameworks and platforms. In enterprise environments, companies don't want isolated agents that can't work together. They need coordination across systems."
The introduction of A2A support and Multivariate Anomaly Detection … create a more intelligent and adaptive system for real-time data analysis. Together, these features empower businesses to act on live signals, automate decisions and prevent issues before they escalate, making them highly valuable.
Stephen CatanzanoAnalyst, Omdia
MCP is a set of open source code created by AI developer Anthropic and launched in November 2024 that enables developers to securely connect the models and data sources that give agents the context they require to properly perform their prescribed tasks.
AWS, Microsoft and Snowflake were among the first data management platform providers to reveal support for MCP. As 2025 evolved, Alation, Databricks, Oracle and Teradata were among many others that followed suit. Confluent unveiled its support for MCP when it introduced Streaming Agents in August 2025.
MCP, however, is limited. While it simplifies developing agents, it does not govern agents once they have been deployed.
While the majority of AI initiatives still don't make it past the pilot stage and into production, some enterprises are successfully developing and deploying agents. To turn those agents into a network capable of interacting and collaborating to be more efficient than when operating in isolation, they need to be connected.
A2A, developed by Google Cloud and released in April 2025, like MCP is a set of open source code. Using A2A, developers and engineers can securely connect and orchestrate agents, enabling collaboration.
By adding support for A2A in Streaming Agents, which serves as an orchestration pane within Confluent that ensures governance and security, Confluent now enables its customers to create collaborative agentic AI ecosystems fed by streaming data.
Perhaps the biggest benefit of adding A2A support is the collaboration it enables among agents fueled by real-time data, according to Kevin Petrie, an analyst at BARC U.S.
He noted that BARC research shows that more than half of all organizations surveyed report that real-time data is critical to their AI initiatives, and many of those are now considering multi-agent networks to address sophisticated use cases.
"Confluent is well-positioned to help smart agents collaborate on a real-time basis, and combining the Apache Kafka messaging system with A2A support is an ideal way to do it," Petrie said. "Agents can publish streams of updates to one another as they receive inputs, make decisions and orchestrate workflows."
Catanzano likewise cited interoperability as a major benefit of adding A2A support.
"The standout is its ability to unlock inter-agent communication and auditability, as this ensures seamless collaboration between AI systems while maintaining governance and traceability, which is critical for enterprise-scale AI ecosystems," he said. "This new standard seems to be getting good traction."
While support for A2A enables Confluent customers to coordinate agentic AI networks, Multivariate Anomaly Detection is designed to help users catch and fix issues in their streaming data pipelines before they reach applications such as agents and affect their performance.
The tool uses machine learning to automatically monitor a set of related metrics to reduce the occurrence of false positives -- many anomaly detection tools monitor just one metric at a time, which can lead to glitches and alert fatigue -- and constantly measures new data points to catch anomalies.
"I like Confluent's approach to ensuring resiliency and uptime," Petrie said. "Agents have many moving parts and interdependencies. ... This big web has many points of failures, and many cascading effects. Multivariate Anomaly Detection seems an effective way to identify and address issues or failures before they cause larger issues."
Meanwhile, from a competitive standpoint, adding A2A support and Multivariate Anomaly Detection help distinguish Confluent from streaming data platforms from tech giants such as AWS and Microsoft, as well as more specialized vendors including Redpanda, according to Catanzano.
"Confluent's additions differentiate it from other streaming data platforms by focusing on intelligent orchestration and real-time adaptability, which are not universally offered by competitors," he said. "These capabilities position Confluent as a leader in connecting AI systems and delivering actionable insights from live data streams."
Informa TechTarget
Looking ahead
As Confluent plots future product development, becoming a platform that enterprises use to develop and deploy large-scale AI systems is a priority, according to Falconer. Toward that end, refining its Real-Time Context Engine, currently in early access, is a focal point.
"It's designed to give agents and AI applications access to fresh, governed, structured context directly from the stream," Falconer said. "The feedback from design partners has been very strong. Our focus is on expanding availability and functionality to address our customers' needs."
Beyond focusing on its own capabilities, Petrie suggested that Confluent partner with AI and machine learning platforms to bolster its current offerings.
"They've partnered with Dataiku and should consider doing the same with vendors such as Domino and H2O.ai," he said.
Likewise, Catanzano advised Confluent to add integrations to expand its ecosystem and broaden its capabilities.
"To continue serving current users and attract new ones, Confluent could expand its AI-driven features by integrating predictive analytics and decision-making tools that leverage historical and real-time data, further enhancing its value for businesses seeking to optimize operations and innovate," he said.
Eric Avidon is a senior news writer for Informa TechTarget and a journalist with more than three decades of experience. He covers analytics and data management.
Informa TechTarget
