Overview

Kavrynt provides a control plane for teams that want to run MCP servers in a repeatable, Kubernetes-native way.

MCP makes tools and context available to AI agents. As teams add more MCP servers, they need a way to track what exists, route traffic predictably, apply policy, and operate the system with familiar platform practices.

Kavrynt starts with that operational foundation.

Problem

Without a control plane, MCP usage can become difficult to manage:

• MCP servers are created by different teams without a shared catalog.
• AI clients need stable endpoints but MCP servers move or change.
• Platform teams need visibility into what is deployed.
• Security teams need a place to add policy, approval, and audit controls.
• Developers need a simple local and Kubernetes install path.

Kavrynt MVP

The MVP focuses on the first useful workflow:

Developer
  -> kavryctl or MCPServer custom resource
  -> Registry
  -> Gateway
  -> MCP server
  -> AI client traffic

Product Shape

Kavrynt has two packaging directions:

Package	Audience	Scope
Open Kavrynt	Developers and platform teams	CLI, Registry API, Gateway, Operator, Helm charts, local install docs.
Kavrynt Cloud	Teams and enterprises	Hosted registry, control UI, SSO, RBAC, audit logs, policy management, usage dashboard, support.

The public docs focus first on the open Kubernetes install path.

Current Status

Kavrynt is an early MVP. It is suitable for local testing, demos, and design feedback. Production hardening work is still planned.