Kubeara MCP server
Commit 28. Okt. 2026Ermöglicht es einem AI-Client, private Infrastruktur zu verwalten, Services zu deployen und GPU/VRAM-Auslastung über eine native MCP-Integration zu überwachen.
Auch bekannt als kubeara/core
- Stars
- 43
- Tools
- —
- Downloads, 30 Tage
- —
- Neueste Version
- —
- Letzter Commit
- 28. Okt. 2026
- Lizenz
- MIT
Was ist Kubeara MCP?
Kubeara ist ein MCP server. Ermöglicht es einem AI-Client, private Infrastruktur zu verwalten, Services zu deployen und GPU/VRAM-Auslastung über eine native MCP-Integration zu überwachen.
Zusammenfassung generiert aus dem README des Projekts
Über Kubeara MCP server
Dieser MCP server ermöglicht es dir, private Infrastruktur und das Deployment von Services direkt über einen AI-Client zu steuern. Entwickler können Tools wie Claude oder Cursor nutzen, um ihre eigenen Server zu kontrollieren, Anwendungen zu deployen und AI-Modelle zu verwalten, ohne manuelle Terminal-Befehle oder SSH verwenden zu müssen. Die Integration funktioniert sowohl mit selbst gehosteten Instanzen als auch mit der verwalteten Control Plane von Kubeara Cloud.
Nutzer können Hardware-Ressourcen in Echtzeit überwachen, insbesondere CPU, RAM, Disk und GPU-Metriken. Der Server bietet spezialisierte Tools für AI-Workloads, darunter einen VRAM-Checker, um den verfügbaren Speicher vor dem Pull von Modellen zu validieren, sowie die Möglichkeit, GPU-Temperatur und Auslastung pro Container zu überwachen. Er unterstützt über 200 Templates für das schnelle Deployment von Datenbanken, Anwendungen und Frameworks wie NestJS, Next.js und React.
Die Sicherheit wird über ein agentenbasiertes Modell gelöst, bei dem der Agent auf deinem Server ausgehende Verbindungen initiiert. Dadurch wird sichergestellt, dass keine SSH-Keys im Klartext in einer Datenbank gespeichert werden. Dieses Setup ist für Teams konzipiert, die hohe Compliance-Anforderungen haben, wie etwa im Healthcare- oder Fintech-Bereich, und ihre Daten auf der eigenen Infrastruktur halten müssen.
Was du damit machen kannst
- 1Lokales Deployment von AI-ModellenEin AI-Client kann AI-Modelle wie Ollama auf deiner privaten Hardware pullen, wechseln und löschen.
- 2Überwachung der GPU-RessourcenEntwickler können die VRAM-Auslastung und GPU-Temperaturen für spezifische Container in Echtzeit prüfen.
- 3Verhindern von fehlgeschlagenen Model-DeploymentsDas System validiert, ob genügend GPU-Speicher verfügbar ist, bevor versucht wird, ein neues Modell zu pullen.
- 4Verwaltung von Multi-Server-InfrastrukturenAgenturen können standardisierte Tech-Stacks über mehrere Client-Server hinweg von einer einzigen Oberfläche aus deployen.
Geschrieben aus dem README des Projekts. Jeder Use Case wurde gegen ein Zitat daraus geprüft.
Automatisierte Checks
Geluften am 15. Nov. 2026Vollständiger Bericht →Noch nicht verifiziert. Ein Server benötigt einen Score von 80, keine kritischen Befunde und genug Daten für unsere Prüfung. Wie die Checks funktionieren · Was du für mehr Punkte verbessern kannst
- Source Repository verknüpft und erreichbarkubeara/core is public and reachablegithub.com/kubeara/core ↗Bestanden
- Paket nennt dieses Repository als QuelleNicht geprüft
- Build-Provenienz (Sigstore-Attestierung)Nicht geprüft
- Open Source Lizenz (OSI)MITspdx.org/licenses/MIT.html ↗Bestanden
- Commit oder Release im letzten JahrLast commit or release 2026-10-28github.com/kubeara/core ↗Bestanden
- OpenSSF ScorecardNicht geprüft
- Package-Name ähnelt keinem populären PackageNicht geprüft
- Keine kritischen Advisories im Paket oder seinen AbhängigkeitenNicht geprüft
- Keine versteckten Anweisungen an das Modell0 tool descriptions and the README read; no override phrases, credential paths or hidden charactersBestanden
- Destruktive Tools sind markiertNicht geprüft
- Remote-Endpoint beantwortet den MCP-HandshakeNicht geprüft
README
Auf GitHub ansehenDeploy anything on your own servers.
Connect a server, click deploy, done.
MIT licensed and free to self-host forever.
Kubeara is an open source platform to deploy and manage any service on your own servers — databases, applications, AI models, and more.
Connect your server. Choose a service. Click deploy. Everything else is handled.
The self-hosted alternative to Coolify, Heroku, Railway, Render, Netlify, and Vercel.
Managed cloud platforms give you convenience but take away control. Your data sits on their servers. Their pricing scales against you. Their outages become your problem.
Kubeara gives you the same one-click deployment experience — on infrastructure you own.
✅ No data leaving your servers.
✅ No per-seat pricing surprises.
✅ No vendor lock-in.
✅ Free forever for self-hosted.
See Kubeara in Action
Two ways to use Kubeara
Self-hosted — free forever
Run Kubeara on your own server.
Full control. MIT licensed.
No feature limits. No time limits.
Kubeara Cloud — managed for you
We run the control plane.
You connect your servers.
Free tier available. No credit card. Try now !
Why Kubeara
Most deployment platforms were built for web applications. Kubeara was built specifically for private infrastructure — databases, AI models, and services that must run on your own servers.
What is built in
| Capability | What it does |
|---|---|
| GPU monitoring | Real-time utilization, VRAM usage, and temperature per container. Not just server-wide — per service. |
| VRAM checker | Validates available GPU memory before every model pull. Prevents failed deployments before they start. |
| AI model management | Pull, switch, and delete AI models from the dashboard. No SSH. No terminal commands. |
| MCP integration | Native MCP server. Manage infrastructure through Claude, Cursor, or any MCP-compatible client. |
| Agent-based security | Agent runs on your server and initiates outbound connections only. No Plain text SSH keys stored in any database. |
| Framework deployment | Framework-specific Dockerfiles for NestJS, Next.js, and React. Works first time without Nixpacks errors. |
| 200+ templates | Every template validated through automated testing including live deployment verification. |
| Free cloud tier | One server free on Kubeara Cloud. No credit card. No time limit. No other platform offers this. |
Who uses Kubeara
| Use case | Why Kubeara |
|---|---|
| Private AI deployment | Run Ollama, Open WebUI, and local LLMs on your own server. Data never leaves your infrastructure. |
| Compliance-driven teams | Healthcare, fintech, and legal teams that cannot send data to third-party AI providers. |
| Agencies and MSPs | Manage multiple client servers from one dashboard. Deploy the same stack across all clients in minutes. |
| Versal, Heroku and Render refugees | Same deployment experience. Your infrastructure. Fraction of the cost. |
| DevOps teams | Replace manual Docker Compose management with a proper control plane and monitoring. |
Quick Start
Self-hosted (free forever)
macOS / Linux / Windows (WSL or Git Bash):
curl -fsSL https://setup.kubeara.dev | sh
Windows PowerShell (Docker Desktop):
irm https://setup.kubeara.dev/install.ps1 | iex
View the install script before running: install.sh
Open dashboard: http://your-server-ip:7935 (or the console URL from your install output).
Manual install:
git clone https://github.com/kubeara/core
cd control-panel
cp .env.example .env
docker compose up -d
Kubeara Cloud (no setup required)
→ Start free at app.kubeara.dev
Free tier — one server, no credit card, no time limit.
After install
- Open the dashboard
- Click Add Server
- Provide server IP and SSH key
- Click Validate and Connect
- Deploy your first service
Why should I use the Cloud version?
Self-hosted is free and always will be. Cloud is for teams who want the same experience without managing the Kubeara control plane themselves.
Your application data never touches our infrastructure on any plan. Both options deploy to servers you connect. We only manage the Kubeara dashboard.
Cloud is the right choice if
- You want automatic Kubeara updates that includes latest security patches, bug fixes and new services
- You want backups handled automatically
- You want email or priority support
- You do not want to maintain the control plane yourself
Pricing
| Plan | Price | Servers |
|---|---|---|
| Free | $0 | 1 |
| Starter | $5/month | 5 |
| Pro | $29/month | 25 |
| Max | $99/month | Unlimited |
| Enterprise | $199/month | Unlimited + SSO |
Features
Free Forever
🚀 Deployment
- One-click deployment from 200+ templates
- Docker Compose and Dockerfile support
- Automatic SSL via Let's Encrypt
- Custom domain configuration
- Automatic container restarts
- Volume and persistent storage management
- Rolling updates with zero downtime
🖥️ Servers
- Connect any server via SSH
- Automatic server validation on connect
- Real-time CPU, RAM, and disk monitoring
- Multi-region server management
- Agent-based security — No plaintext SSH key storage
📊 Real-Time Metrics
- Monitor CPU, RAM, and disk usage
- Monitor GPU utilization and temperature
- Track VRAM usage per service
- View real-time resource usage across your infrastructure
🛡️ Resource Validation
- Validate available resources before deployment
- Prevent failed deployments caused by insufficient resources
- Check CPU, RAM, GPU, and VRAM availability
🤖 AI & Models
- One-click Ollama and AI model deployment
- VRAM validation before every model pull
- GPU utilization per container
- GPU temperature monitoring
- VRAM usage per service
📋 Activity Tracking
- Keep track of deployment and infrastructure activities
- Monitor important changes across your servers and services
🛠️ Developer Experience
- Native MCP server integration
- Real-time deployment logs
- NestJS and Next.js framework support
- API key management
- Team member management
Enterprise
- 🔑 SSO / SAML / OIDC
- 📂 LDAP / Active Directory
- 📋 Full audit logs with export
- 🛡️ Advanced RBAC — custom roles
- 📄 Compliance documentation
- 👥 Unlimited team members
- 🤝 SLA guarantee
- 💬 Priority support
- 🏷️ White-label rights
→ Enterprise pricing
→ Request enterprise trial
Development workflow
Kubeara uses Conventional Commits and Release Please for automated, semver-based releases.
Feature branch
↓
Pull request (CI: lint, build, test)
↓
Merge into main
↓
Release Please opens/updates a Release PR
↓
Review Release PR (version + CHANGELOG.md)
↓
Merge Release PR
↓
Git tag + GitHub Release published
↓
Users install the new version
Branching
- Create a feature branch from
main(for examplefeature/deployment-logs). - Open a pull request into
main. - Ensure CI passes (lint, build, template tests).
- Merge after review.
Merging to main does not publish a release immediately. Release Please accumulates commits and opens a Release PR when a version bump is warranted.
Conventional commits
Use these types in PR titles or squash-merge commit messages:
| Type | Release impact | Example |
|---|---|---|
feat |
Minor (1.0.0 → 1.1.0) |
feat: add deployment log streaming |
fix |
Patch (1.0.0 → 1.0.1) |
fix: resolve docker compose parser error |
feat! or BREAKING CHANGE: |
Major (1.0.0 → 2.0.0) |
feat!: redesign deployment API |
docs |
Changelog only | docs: update install guide |
refactor |
Changelog only | refactor: simplify agent install flow |
perf |
Performance section | perf: reduce deployment status polling |
test |
No release | test: add template compose validation |
build / ci / chore / style |
No release | chore: update dependencies |
Breaking change examples:
feat!: remove legacy deploy endpoint
feat: add new deploy API
BREAKING CHANGE: removed /api/v1/deployments/legacy
Releases
- Release PR: created/updated by Release Please on pushes to
main. - On merge: bumps
package.json, updatesCHANGELOG.md, creates a git tag (for examplev1.2.0), and publishes a GitHub Release. - Tracked version:
.release-please-manifest.json.
See also release-please-config.json and .github/workflows/release-please.yml.
Security
Security model
Kubeara uses an agent-based architecture designed to eliminate the attack surface that caused the January 2026 Coolify vulnerabilities.
How it works:
- Agent runs on your server
- Agent initiates outbound connections only
- No Plain text SSH keys stored in any database
- No persistent access to your server after initial bootstrap
- All agent communication encrypted in transit
- This project is tested with BrowserStack
Reporting a vulnerability
Please do not open a public GitHub issue for security vulnerabilities.
Email: bhushan.lilapra@kubeara.dev
We will respond within 48 hours. We follow responsible disclosure. We credit researchers in our changelog.
Supported versions
| Version | Supported |
|---|---|
| Latest | ✅ |
| Previous minor | ✅ |
| Older versions | ❌ |
We recommend always running the latest version.
Known security measures
✅ Plain text SSH keys never stored in database
✅ All secrets encrypted at rest
✅ Agent outbound connections only
✅ No persistent SSH access after bootstrap
✅ Rate limiting on all API endpoints
✅ Automatic security updates via Docker
✅ HTTPS enforced on all connections
Kubeara MCP Fragen
Kann ich meine Infrastruktur mit Claude oder Cursor verwalten?
Ja, Kubeara enthält einen nativen MCP server, mit dem du deine Infrastruktur über Claude, Cursor oder jeden anderen MCP-kompatiblen Client verwalten kannst.
Wie regelt Kubeara die Sicherheit für Serververbindungen?
Es nutzt eine agentenbasierte Architektur, bei der der Agent auf deinem Server nur ausgehende Verbindungen initiiert. So wird sichergestellt, dass keine SSH-Keys im Klartext in einer Datenbank gespeichert werden.