Solo.io Donates Kagent to CNCF, Introduces MCP Gateway
Roughly two weeks after introducing Kagent, an open source framework for AI agents run in Kubernetes, the project’s creator donated it to the Cloud Native Computing Foundation (CNCF).
Solo.io announced the move earlier in April at KubeCon + CloudNativeCon Europe, in London.
The move is a reflection of Kagent’s status as an instant hit with developers. “In order for a project to be added to the [CNCF] landscape, you have to have 300 GitHub stars,” Lin Sun, Solo.io.’s senior director of open source, told The New Stack. “Actually we hit it within a week when we open sourced Kagent.”
Solo.io made other announcements about its open source projects at KubeCon in London:
Kgateway, an open sourced version of Gloo Gateway that serves as an Envoy data plane proxy, has been accepted as a CNCF sandbox project. Kgateway was donated to the CNCF during November’s KubeCon + CloudNativeCon North America in Salt Lake City.
MCP Gateway, a new open source tool designed to work with the Model Context Protocol (MCP) developed by Anthropic, was also announced.
MCP Gateway can “federate all the tool servers to a single endpoint,” said Solo.io Co-founder and CEO Idit Levine in a KubeCon keynote. And then, it is “seamlessly going to multiplex all the traffic to the upstream tool.”
The MCP Gateway project has a “tight integration with Kgateway,” Sun said. “You can think about Kgateway as the control plane to programming the MCP gateway proxies.”
Two new tools for Ambient Mesh, an evolution of Istio service mesh that doesn’t require sidecars, were announced into general availability.
First is an automated zero-downtime migration tool that eases the way for users to migrate from sidecars to Ambrient. Users can run the tool against their Kubernetes cluster, Sun said, and learn “what are the things in your cluster that are ready to move to Ambient, and what’s not, so you can[work] on the suggestion of what’s not ready.
She added, “And then run the tool again, until the tool tells you, ‘Hey, you’re ready to move to Ambient. These are the steps you should take to move to Ambient.’”
And secondly is a cost-analysis tool that looks at the costs of CPU and memory and estimates the potential savings of migrating from sidecars to Ambient Mesh.
“We’re really excited because we had some early adopters of Ambient who run through the cost and analysis from their end,” Sun said. “They’re talking about saving. We’re not talking about thousands of dollars. We’re talking about millions of dollars of savings.”
Kagent’s Fast-Growing Community
The community around Kagent is growing rapidly, Sun said. When she spoke to TNS during KubeCon in London, she noted that “25% of our PRs [pull requests] merged in Kagent since we launched actually are contributed by people outside of our company.”
She attributed the activity to “this excitement around agentic AI and solving cloud native operation challenges. Also, people want to learn this space. [It] really motivates a lot of people learning and contributing to Kagent.”
As a technical oversight committee member at the CNCF, Sun anticipated a short wait until Kagent joins Kgateway in the foundation’s sandbox. Kgateway, she noted, only took a couple of months from donation to acceptance as a sandbox project.
“Surprisingly, our queue is very short,” she said. “So I would say, based on the same speed, we’re looking at Kagent becoming CNCF projects in the same timeframe. Like May, June-ish.”
Solo’s top goals for 2025, Sun said, include growing its open source contribution and community. According to Devstats, which tracks contributions to CNCF projects, Solo.io was the ninth largest corporate contributor to open source projects in the CNCF landscape (excluding independents) the day she spoke to TNS during KubeCon.
She’s proud of that achievement, partly for how Solo.io has performed as an open source contributor alongside giants like Google and IBM.
“You’re looking at companies who hire tens of thousands of employees and their contribution on the same charts. As a small company like Solo, we only have 150 people, right around that, so that shows how much commitment we have around open source.”
Preparing for the Agentic AI Future
Solo.io is betting a lot on ambient mesh, feeling it’s a more efficient architecture. And one that, as Sun put it, can “really solve the long-term pain users have around service mesh.”
As AI — and in particular agentic AI — grows in adoption, Sun believes Solo.io is well-positioned to solve the problems it brings for organizations that work at scale.
“We’ve seen tremendous interest around this area, and we do believe the gateway and mesh empowered for AI is going to be very, very relevant.”
She added, “What Solo is really focusing on is, how do we solve the next challenges as people trying to run agentic AI workload in Kubernetes, how do we make it easy for users? How do we solve the same connectivity problem and security problem and observability problem, but particularly for agentic AI workloads? Because those have specific challenges on top of traditional microservices.
“A lot of problems they could potentially ignore with Kubernetes and workloads today, it becomes more important with agent AI workloads, as they want to get the costs under control and be more efficient.”
