Deployment and Operations

This section explains how to host and operate MCP Runtime, using our public platform at platform.mcpruntime.org as the reference deployment. We walk through the infrastructure choices, cluster preparation, platform installation, identity integration, updates, backups, and verification that make up a complete deployment.

The public reference uses K3s for its Kubernetes cluster and a separate Docker/Caddy VM for Keycloak. The guides identify those choices where they affect commands and configuration. You can apply the same deployment lifecycle while selecting another Kubernetes distribution, registry, ingress controller, storage system, or compatible identity provider.

To use the existing public platform, follow the Public Platform Walkthrough to try user and team paths. To try platform setup and administration, install it in your own environment. The pages in this section are for operators hosting their own platform or learning how the public reference is operated.

Installation

Start here when hosting your own platform. These guides apply to managed and self-managed Kubernetes clusters.

1. Deployment Options: choose a Kubernetes distribution, registry, ingress, TLS, and storage model.
2. Cluster Requirements: check node image pulls, DNS, ingress, certificates, and storage before installing.
3. Platform Installation: install the CLI, configure the platform, check its health, and publish your first server.

Public reference

Follow the public platform's worked example to see how those choices fit together in a complete deployment.

1. Cluster Provisioning: build the example node topology, configure cluster access, and prepare ingress.
2. Deployment Guide: configure Runtime and its external identity provider, install and update the platform, back up and recover data, and verify the deployment.

The provisioning guide hands off to the installation guides once the cluster is ready. Adapt the example's domains, node layout, credentials, and backup ownership to your environment before running the reference scripts.

Operations

Return to these guides after installation for component placement, runtime behavior, service maintenance, and incident diagnosis.

• Runtime Operations: server resources, reconciliation, routing, and rollouts.
• Namespaces: component placement, team boundaries, service DNS, and credential ownership.
• Services and Observability: platform services, gateway enforcement, audit events, metrics, logs, and traces.
• Follow a Request: correlate a client response's X-Trace-ID with traces, logs and metrics.
• Troubleshooting: diagnosis of installation, access, registry, analytics, and cluster failures.

Use Follow a Request to correlate client payloads, gateway decisions, traces, metrics, logs and read-only Kubernetes diagnostics.