Introduction

What Kubeara is, who it is for, and how one-click deploys, security, and MCP fit together.

Kubeara is a self-hosted infrastructure platform for teams that want databases, caches, apps, and AI models running on servers they own — without a week of Docker, nginx, and certbot work.

Connect a server over SSH, pick a stack, and deploy. Kubeara configures Docker, firewall rules, TLS, health checks, and volumes automatically.

Getting started

Adopt Kubeara step by step: Cloud or self-host, connect a server, deploy a service.

Service catalog

Browse 200+ templates. SSL, health checks, and volumes are configured for you.

MCP

Manage servers and deploys from Claude, Cursor, ChatGPT, and VS Code.

Security

Outbound-only agents, encrypted credentials, and audit-ready controls.

What Kubeara does

Kubeara is the control panel for private infrastructure. Workloads stay on your machines. The panel coordinates connection, deploy, logs, and updates.

Core capability areas include:

Server management — connect Ubuntu hosts with SSH, then let an outbound-only agent keep the session
One-click services — PostgreSQL, Redis, Kafka, Grafana, NestJS, Next.js, Ollama, and 200+ other stacks
Deploy automation — Docker, TLS via Let's Encrypt, reverse proxy, health checks, and persistent volumes
GPU and AI — per-container GPU metrics and VRAM checks before model pulls
Developer access — HTTP API plus native MCP for AI clients
Team operations — logs, restarts, updates, and (on Enterprise) SSO, LDAP, and audit logs

Problems Kubeara addresses

Most teams still assemble a PaaS by hand: install Docker, write Compose files, open ports, issue certificates, wire a reverse proxy, then debug why nothing starts. That work does not ship product.

Typical pain:

Days spent on infrastructure before the first database is reachable
SSH keys copied into dashboards or tickets in plain text
Inbound management ports left open so a control panel can “reach in”
AI assistants that each need their own ad-hoc SSH or API hack
Failed LLM pulls because the host did not have enough VRAM

Kubeara is built so you connect a server and deploy, instead of maintaining a private Heroku clone.

Who Kubeara is for

Engineers who need Postgres, Redis, or an app on a VPS today
Platform and DevOps teams standardizing deploys on Hetzner, bare metal, or a private VPC
Teams running local or private AI (Ollama, Open WebUI) with GPU visibility
Founders who want MIT-licensed self-hosting with an optional managed Cloud control plane

How Kubeara differs from general-purpose PaaS

Heroku, Render, and AWS are general-purpose clouds. Kubeara is a control panel for servers you already have.

Topic	Typical PaaS	Kubeara
Where workloads run	Vendor region	Your VPS, metal, or VPC
Self-host the panel	Rarely	MIT licensed, free forever
SSH credentials	Often stored or not used	Never stored in plain text
Agent direction	Inbound to the host	Outbound-only after connect
GPU / VRAM	Extra product	Built in
AI assistants	Separate integrations	Native MCP

Supported connections

Hosts — Ubuntu 22.04+ with SSH from the machine running Kubeara (Cloud) or from your self-hosted panel
Workloads — Docker-based catalog services, Compose, and framework Dockerfiles (NestJS, Next.js)
AI clients — Claude Desktop, Cursor, ChatGPT, VS Code / Copilot via MCP
Identity (Enterprise) — SSO and LDAP

Product terminology

Term	Meaning in Kubeara
Cloud	Managed control panel at app.kubeara.dev. You still connect your servers.
Self-hosted	You run the control panel (install script or Docker Compose). MIT licensed.
Server	A connected Ubuntu host (name, host, username, SSH credentials).
Agent	Process on the server that talks out to Kubeara. No inbound management port required for Kubeara.
Service	A catalog stack or custom app you deploy onto a server.
Deploy	Job that pulls images, configures TLS/proxy/volumes, and streams logs.
MCP	Model Context Protocol — how AI tools call Kubeara with one permission model.

Next steps

Getting started — Cloud or self-host, then first deploy
Requirements — Ubuntu, SSH, and network
Self-hosting — curl \| sh or Docker Compose
FAQ — licensing, privacy, and catalog questions