FEA: MCP Registry in Kubernetes, MVP-1

[dmartinol]

See #1641 for detailed plan.
This tasks is split into multiple sub-tasks for more granular tracking and incremental delivery.

Overview

This feature defines a Kubernetes-native solution for managing MCP (Model Context Protocol) server registries using the ToolHive software.

The solution provides a declarative approach to registry management through Custom Resource Definitions (CRDs) and includes a REST API for programmatic access.

This MVP-1 implementation can be extended with additional functionalities in future phases, as there are no existing similar deployments in the upstream MCP specifications.

Goals and Objectives

Primary Goals

• Runtime Migration: Migrate existing local registry functionalities to Kubernetes cluster environment
• Kubernetes-native: Fully integrated with Kubernetes using operators and CRDs
• Declarative Management: Registry configuration through Kubernetes manifests to simplify DevOps operations
• Multi-source Support: Extensible architecture for various registry data sources
  • MVP-1 focuses on ConfigMap-based data sources
• Format Flexibility: Support for both upstream and [ToolHive] (https://github.com/stacklok/toolhive/blob/main/pkg/registry/data/schema.json) registry formats
  • MVP-1 covers only toolhive format but introduces the required interfaces to work with format conversions later
• REST API Access: HTTP endpoints for programmatic registry interaction, compliant with the upstream schema
• Server Lifecycle Management: Automated watching of MCP servers using label-based associations
• Automated Operations: Automated sync capabilities

Non-Goals (Future Phases)

Ordered by decreasing priority:

• thv CLI integration changes
• Support for other source types: URL, git, registry
• Authentication/authorization
• Advanced deployment configurations (replicas, resources, expose route, ...)
• Server build workflows (for non-docker packages like npm and python)
• Server validation workflows (including MCP security checks)
• Registry federation and aggregation
• Registry publishing/contribution workflows
• Catalog definition with configurable validation policies
• Advanced telemetry and tracing capabilities