Solo.io has released ‘Kagent,’ a new open source framework for DevOps pros and platform engineers to run AI agents in cloud native environments.
It works by providing a foundation on which teams can build internal automation platforms inside Kubernetes environments. Agents built using Kagent will be well placed to automate tasks such as configuration, troubleshooting, observability, and network security, Solo.io said.
Kagent integrates with other cloud native tools through an architecture built on the Model Context Protocol (MCP). This means teams can run these tools at scale without needing expert knowledge in every area of the cloud native ecosystem.
Solo.io said that, with the tool, teams can offload heavy lifting to agentic infrastructure in order to concentrate on more valuable tasks.
“Kubernetes is already the go-to platform for predictive analytics, MLOps, and inferencing, and now it provides the modern architecture to deploy AI agents dynamically,” Keith Babo, CPO at Solo.io, said.
“Kagent will enable Kubernetes users to run agentic AI without the huge learning curve and operational challenges of building an agentic AI infrastructure stack from scratch,” Babo added.
How does Kagent work?
According to Lin Sun, the director of open source at Solo.io, the idea for Kagent came about on the back of a productivity drive internally which forced the company to ask questions about how agentic AI could be used to create efficiencies.
Sun told ITPro the firm asked whether it could leverage agentic AI to tackle the common issues faced by support engineers and customers of the platform, or whether it could build the expertise from its best engineers into agents for other staff to use.
“It’s kind of our way of cloning some of our top engineers – instead of having them be distracted to work with customers – so that they can focus more on coding and innovating,” Sun said.
Now, the platform is available to solo.io’s customers, consisting of three different layers needed to implement agentic architectures on cloud native infrastructure.
There’s a tool layer that gives users access to pre-defined functions that AI agents can use like expert knowledge bases, availability and performance metrics, application deployment and lifecycle controls, platform admin tools, and more.
Then there are the agents themselves, which can be used within the platform and configured to undertake multi-stage activities like canary deployments for new application versions, establishing zero trust policies, or debugging service availability problems.
The final part of Kagent is a declarative framework layer made up of an API and controller for building and running agents.
Agents are all the rage
Solo.io is the latest company to mark its foray into the world of agentic AI, with all the major tech firms having released agentic offerings and tooling over the last few months.
Some recent developments include Google Cloud’s announcement of new sovereign cloud services in the UK which will include UK data residency specifically for Google Agentspace, the firm’s agent platform.
Cisco also revealed its own agentic offering recently that came as a suite of new tools targeting both customers and employees in customer services and device collaboration capacities.
The benefits of agents are clear, with workers already citing positive results. For example, research from Pegasystems found that nearly 60% of staff in the UK and US are using agents daily, with many saying the automation of tedious work was the main benefit.
Agents are set to be a big money spinner for tech firms, with analysis from venture capital firm Sequoia Capital claiming that firms could soon be tapping into the multi-trillion dollar global services market.
