Documentation

Platform Documentation

Everything you need to understand, evaluate, and start using mctl

Overview

What is mctl?

mctl is a self-service Kubernetes platform that automates the full delivery lifecycle for product teams. It combines GitOps (ArgoCD), secrets management (HashiCorp Vault), team isolation (RBAC, namespaces, network policies), and AI-native tooling (MCP server) into a single control plane.

Instead of assembling and maintaining these pieces yourself, mctl provides them pre-integrated and production-ready. Teams get isolated workspaces, pick from a service catalog, and deploy via git push, a developer portal, REST API, or natural language through any MCP-compatible AI client.

Everything is automated end-to-end: from namespace provisioning and RBAC setup to CI/CD pipelines, TLS certificates, database provisioning, and domain management. No kubectl, no manual YAML, no tickets.

Use Cases

Who is it for?

Startups (5-50 engineers)

You have multiple teams shipping features but no dedicated DevOps hire. mctl gives every team its own isolated namespace with secrets, TLS, and CI/CD from day one — so engineers deploy themselves instead of waiting on infrastructure.

B2B SaaS Companies

You need per-customer or per-environment isolation with consistent security and governance. mctl provisions tenants with their own namespaces, resource quotas, and network policies. Onboard a new customer environment in minutes, not days.

Platform Teams

Already building an internal developer platform? mctl provides the foundation: GitOps engine, service catalog, secrets integration, and self-service workflows. Skip 6 months of integration work and build your customizations on top.

AI-first Teams

You want infrastructure managed through natural language. mctl exposes 20+ MCP tools that let Claude, Cursor, VS Code Copilot, or any MCP client deploy services, manage previews, provision databases, and monitor resources through conversation.

Architecture

How it works

Developer / AI

Portal, API, MCP

mctl-api

Control Plane

Argo Workflows

Task Execution

mctl-gitops

Source of Truth

ArgoCD → K8s

Reconcile & Deploy

Every action — whether triggered by a git push, a portal click, an API call, or an AI prompt — flows through mctl-api, which orchestrates Argo Workflows to make changes in the mctl-gitops repository. ArgoCD watches the repo and reconciles the cluster state. This means every change is a git commit, every deploy is traceable, and rollbacks are a one-line revert.

mctl-agent: self-healing loop

A separate component watches the cluster for anomalies (crash loops, resource pressure, misconfiguration). When it detects an issue, it diagnoses the root cause using Claude API and opens a pull request to mctl-gitops with the fix. The change goes through the same GitOps pipeline — reviewed, merged, deployed.

Components

Platform components

mctl-api

Control Plane

Go REST API and MCP server. Receives commands from the portal, CLI, AI clients, or direct HTTP calls. Authenticates via GitHub OAuth or Dex SSO. Routes operations to Argo Workflows for execution. Exposes 20+ MCP tools for AI-driven management.

mctl-gitops

Source of Truth

The GitOps repository. Contains Helm charts, ArgoCD application definitions, Kubernetes manifests, and tenant configurations. Every platform change is a git commit here. ArgoCD watches this repo and reconciles cluster state continuously.

mctl-portal

Developer Portal

Backstage-based developer portal. Provides a service catalog, software templates, deploy UI, and team management. Developers browse available templates, configure services, and track deployments through a web interface.

mctl-agent

Self-Healing

Watches the cluster for issues: crash loops, OOM kills, failed probes, resource exhaustion. Diagnoses problems using Claude API, then opens pull requests to mctl-gitops with targeted fixes. Humans review, merge, ArgoCD deploys.

Interfaces

Four ways to interact

Portal

app.mctl.ai

Web UI built on Backstage. Browse the service catalog, create services from templates, view deploy history, manage team members, and check logs.

AI / MCP

api.mctl.ai/mcp

Natural language interface via MCP protocol. Connect Claude, Cursor, VS Code, Copilot, or any MCP client. Deploy, rollback, preview, provision databases — all through conversation.

REST API

api.mctl.ai

Direct HTTP access to every platform operation. Use from CI/CD pipelines, scripts, or custom tooling. Authenticate with GitHub token or Dex JWT.

GitOps

push to repo

Push code to your service repo. CI builds the image, updates the GitOps repo, and ArgoCD syncs to the cluster. Zero-touch delivery from commit to production.

Quick Start

Getting started

Developer

Sign in with GitHub at mctl.ai. Pick a template from the service catalog. Configure your service (name, database, domain). Deploy — your namespace, CI/CD pipeline, and TLS certificate are provisioned automatically.

Platform Operator admin

Create a tenant via the API or portal. Assign namespaces, set resource quotas, configure network policies. Manage services and environments through the REST API or ArgoCD directly. Requires admin-level access.

AI User

Go to docs.mctl.ai/mcp/connecting, sign in with GitHub and add MCTL to your AI client. Start prompting: "deploy my-app to staging", "show service status", "create a preview for PR #42".