
Solo.io today made available an enterprise edition of an open source agent gateway it previously donated to the Linux Foundation, in addition to launching a consulting service that makes engineers with expertise in implementing the Model Context Protocol (MCP) available to IT teams deploying AI agents.
Collectively, Solo Enterprise for agentgateway and Solo Labs for MCP service provide a holistic approach for providing the connectivity, security and observability capabilities that AI agents running in production environments will require, says Keith Babo, chief product officer for Solo.io
Launched earlier this year, the agentgateway project provides a data plane for natively implementing agentic AI protocols such as the Model Context Protocol (MCP), originally developed by Anthropic and the Agent2Agent (A2A) framework developed by Google that is now being advanced under the auspices of the Linux Foundation alongside the agentgateway project.
That approach provides a platform that provides the ability to virtualize and federate MCP tool servers to provide a single access and governance point for MCP interactions. It is specifically designed from the ground up to manage agentic AI connectivity and security versus trying to extend an existing gateway, notes Babo. In fact, when compared to existing alternative legacy gateways, Solo.io claims agentgateway offers 300x better memory utilization, 35x higher throughput, and more than a 100x reduction in latency.
At the same time, Solo Enterprise provides authentication and authorization controls to any service calling to the agentgateway, which provides an advanced implementation of secure elicitation that is seamlessly integrated with identity provider and identity access management (IAM) infrastructure, notes Babo.
Other capabilities include onboarding and registration tools, server fingerprinting, versioning and runtime policy to protect against cyberattacks involving tool poisoning, rug-pulls, tool shadowing, and naming collisions before they start.
Agentgateway also supports automatic, secure token exchange for agent and tool interactions to limit permissions to exactly what is required for an agent and tool to act on a user’s behalf. End-to-end audit trails for user interactions with agents and tools are cryptographically verifiable to support attestation and access reporting controls.
Agentgateway is also designed to drop into existing agentic frameworks, integrated development environments (IDEs) or applications as a virtualized MCP tool server requiring no refactors or code changes. Deploy as an Ambient service mesh waypoint for transparent traffic redirection, tool-server sandboxing, and automatic gateway injection. IT teams can also choose to import OpenAPI specifications and expose them as governed MCP tools.
It is, of course, still early days when it comes to deploying MCP servers or an A2A framework, so IT teams would be well-advised to start small, says Babo. Not all MCP servers, for example, are equally mature, and in many cases a gateway will provide crucial security and governance capabilities that can be more consistently applied, he added.
In the meantime, however, IT teams should start to starting thinking through exactly how the way application connectivity will need to fundamentally change and evolve in the age of agentic AI.