Mirantis launches self-hosted AI agent platform for Kubernetes
Mirantis announced general availability of the Mirantis Agent Platform on November 10, 2026, during KubeCon + CloudNativeCon North America in Salt Lake City, positioning the release as a production-ready tool for enterprises running AI agents at scale with enforceable guardrails.
Self-Hosted Infrastructure and Deployment
The platform runs on an organization's own Kubernetes infrastructure, installing via a single Helm chart. This architecture lets teams keep agents within their own environment rather than routing them through third-party cloud services. Mirantis says the platform is designed to let organizations run agents "where you choose, scales as fast as you need, and is held to the policies you set."
Governance and Control Mechanisms
The Agent Platform includes policy-based governance, configurable spending limits, full audit trails, network-access cutoff capabilities, sandboxed execution, and human approval workflows for proposed remediation. These controls address operational and security concerns that enterprises raise when deploying autonomous systems in production environments.
Mirantis is bundling the platform with a Forward Deployed Engineering (FDE) service, under which senior Mirantis engineers work directly with customer teams to design, integrate, and operationalize custom agents tailored to specific workflows.
Prism Agent and Use Case
The platform includes a named agent called Prism Agent, which is designed to identify cluster problems and propose fixes for human review before execution. This reflects a pattern in production AI agent deployment: autonomous diagnosis paired with human decision gates rather than fully autonomous remediation.
Market Context
The announcement comes as enterprises signal growing interest in agent tooling that runs on-premises. Mirantis exhibited the platform at Booth 232 during the conference, which ran November 9–12, 2026. The timing reflects the broader shift toward agentic operational infrastructure frameworks in cloud-native environments, where teams need oversight mechanisms that align with existing governance structures and compliance requirements.
The general availability release indicates the platform has moved beyond beta or preview status and is ready for production workloads, marking a concrete product milestone in the infrastructure layer of the agent economy.
