Agent fleets drag the messaging layer from afterthought to AI’s coordination core

Data center telemetry, AI workloads and agent fleets are generating more machine-to-machine traffic than ever before, highlighting the limitations of traditional messaging systems designed for more stable demand. This shift has put the spotlight on the open-source Neural Autonomic Transport System (NATS), a messaging system that enables applications and services to exchange messages based on topics.

NATS was created in 2010 and became part of the Cloud Native Computing Foundation in November 2026. Its simplicity makes it well-suited for AI workloads, according to Andrew Connolly, the project maintainer at Synadia Communications Inc., a provider of managed NATS services. With AI tools like Lovable capable of turning natural language prompts into functional web applications, the messaging layer must maintain continuous updates as sandboxes appear and disappear rapidly.

Connolly explained that "everything is really ephemeral" in the current landscape, noting that "if you take the Lovable example, connections coming and going really fast, tons of churn, tons of sandboxes, all these things spinning up, spinning down." Not every technology can cope with such high levels of churn. Both Connolly and Waldemar Quevedo, another project maintainer at Synadia, discussed NATS's capabilities during an exclusive broadcast at KubeCon + CloudNativeCon NA.

They highlighted how NATS manages traffic from short-lived sandboxes, carries data center telemetry, and supports AI agents in coordinating work. NATS has expanded from single-cluster deployments to a system spanning data centers, regions, and clouds, including a recent instance where an AI app builder rebuilt its real-time service on NATS to handle millions of subscriptions.

NATS can also route messages across fleets of AI agents, using its JetStream persistence engine and key-value store to preserve the state necessary for agents to delegate tasks effectively. Connolly emphasized that "in some sense, agentic systems are just microservices systems, but at a bigger scale," concluding that NATS is well-positioned to support the growing complexity of AI agent-based systems.