As enterprises race to deploy AI agents in production, a critical gap has emerged in the infrastructure stack. While traditional cloud-native workloads have mature security, observability, and governance frameworks, AI agents operate in what Christian Posta, VP and Global Field CTO at Solo.io, calls “the Wild West” of trust and connectivity.

Solo.io’s latest announcement addresses this gap head-on with Agent Gateway, an open source data plane specifically optimized for agent AI connectivity. This isn’t just another proxy—it’s a purpose-built solution for the unique challenges of agentic AI communication patterns.

The Agent Communication Challenge

Unlike traditional microservices that follow stateless request-response patterns, AI agents communicate through stateful protocols that maintain context and sessions. These agents need to talk to Large Language Models (LLMs), interact with tools through protocols like Model Context Protocol (MCP), and communicate with other agents using emerging standards like Agent-to-Agent (A2A) protocol.

“Those interaction patterns are stateful in order to maintain context and sessions. As you begin to scale, managing those sessions becomes resource-intensive,” explains Posta. “This is why Solo.io built Agent Gateway in Rust—performance is non-negotiable when handling resource-intensive, stateful communications at scale.”

Building on Proven Foundations

Solo.io’s approach demonstrates mature thinking about enterprise AI infrastructure. Rather than reinventing the wheel, Agent Gateway integrates seamlessly with established cloud-native technologies including Istio, Ztunnel, and Solo.io’s own KGateway and Kagent projects. This creates what they term an “agent mesh architecture”—bringing the same level of connectivity sophistication that exists for traditional workloads to AI applications.

Keith Babo, Chief Product Officer at Solo.io, emphasizes their philosophy: “Every opportunity we have to build on that foundation, we take. We’re essentially just filling the gaps in the open ecosystem.”

Security and Observability: From Wild West to Enterprise Grade

The security implications of autonomous agents are profound. Unlike deterministic traditional applications, agents are probabilistic—they won’t necessarily take the same steps every time. This unpredictability requires new approaches to boundaries, authorization, and audit trails.

Agent Gateway addresses these concerns by providing transparent security and observability across all agent interactions, regardless of the underlying AI frameworks or tools being used. It supports fine-grained authorization, cryptographically verifiable agent identities, and comprehensive tracing across the entire agent communication stack.

“If you really want to explain or debug how that agent arrived at a given outcome, you’re going to rely on network tracing, metrics, and logging to aggregate all that information,” Babo explains. This unified observability approach is crucial for enterprises that need to debug, improve, and explain agent behavior in production.

The Platform Team Dilemma

Enterprise platform teams face a familiar challenge with AI workloads—they’ll inevitably need to support these systems, but they’re often not included in the development process. Solo.io has seen this movie before with earlier technology shifts.

“We know we need to evolve there,” says Babo, noting that platform teams are taking both reactive and proactive approaches. Reactively, they’re preparing for the inevitable arrival of AI workloads on their platforms. Proactively, they’re using tools like Kagent to build custom agents that handle routine platform tasks, freeing engineers to focus on higher-value work.

Open Source First, Commercial Value Second

Solo.io’s commitment to open source runs deep—they’re the ninth-largest contributor to the Cloud Native Computing Foundation despite being a relatively small company. This philosophy extends to Agent Gateway and their broader AI infrastructure approach.

“Everything we’re building is based on an open source foundation and open standards,” Babo emphasizes. The company’s commercial offerings extend these open-source projects with enterprise features and support, but the core technology remains community-driven.

Looking Ahead

The response to Solo.io’s AI infrastructure projects has exceeded expectations, particularly around Kagent for Kubernetes-based agent deployment. The company hints at significant partnerships and announcements coming in the next six months, focusing on creating what they hope will become the leading agent framework for Kubernetes.

As enterprises move beyond AI experimentation toward production deployments, infrastructure considerations become paramount. Solo.io’s Agent Gateway represents a crucial stepping stone—bringing the same level of infrastructure maturity that enabled cloud-native adoption to the emerging world of agentic AI.

For organizations serious about deploying AI agents at scale, the question isn’t whether they need this level of infrastructure sophistication, but how quickly they can implement it before their AI initiatives outgrow ad-hoc solutions.