Krumware bridges Kubernetes's developer experience gap with new MCP server

Most development teams still juggle multiple tools to push code to production. Infrastructure teams understand their Kubernetes clusters, developers understand their applications, but a meaningful gap separates the two—and throwing more tools at the problem hasn't solved it.

Krumware is approaching this friction point differently. Rather than adding another layer on top of Kubernetes, the company is focusing on what it calls a "develop like you deploy" methodology through Epinio, an open-source application development engine that runs natively within Kubernetes clusters. The latest addition—an Epinio MCP (Model Context Protocol) server—standardizes how AI agents and developers interact with the platform while respecting existing organizational guardrails.

Shrinking the local-to-production divide#

The core problem is environmental mismatch. A developer's local machine might run Kubernetes through Rancher Desktop or minikube, but that setup often differs from the production cluster—different security rules, compliance constraints, and networking configurations. This gap forces problems downstream.

Epinio eliminates this friction through a single command workflow. An epinio push takes source code from a local file, automatically detects the language and runtime using Cloud Buildpacks, builds a container image, deploys it into the cluster, and surfaces a live URL—all within seconds. Because Epinio runs natively inside Kubernetes everywhere, the entire build process inherits production's security rules and compliance controls locally. Failures surface earlier, caught by the right guardrails, rather than surprising teams in production.

The experience mirrors Cloud Foundry's simplicity on legacy infrastructure: developers don't need to understand Kubernetes internals. As Krumware CEO Colin Griffin notes, the goal is that "you don't need to know what's under the hood."

AI and platform engineering converging#

The AI boom is shining a light on a deeper organizational need. Many enterprises have invested heavily in platform tools—some spending hundreds of thousands on Backstage instances—without clearly understanding how developers actually use Kubernetes. Adding AI-assisted tools atop weak foundations often widens the gap between platform-ready and platform-unprepared teams.

Griffin, who co-authored the CNCF's Platform Engineering Maturity Model, frames "AI readiness" as inseparable from platform engineering readiness. Security, observability, networking, tool interoperability, and self-service workloads—the foundations of platform engineering—are prerequisites for safely deploying AI agents. Platform engineering becomes the enablement layer for everything that follows.