Run apps, databases, AI and Kubernetes on servers you own
SigmaHub is a managed cloud control plane for your own machines — Git deploys, managed databases, S3 storage, GPU/LLM inference, monitoring, disaster recovery and Kubernetes, from one dashboard. You pay for the platform, never for the servers.
Free for up to 3 servers · €5 per connected server after that · every feature included
Overview
production · 8 servers
Connected servers
Monthly cost
Deploys / 7d
Fleet uptime
Deploys & requests
7d
webshop
app · fra-1
llama-3-70b
gpu · fsn-1
postgres-prod
db · fsn-1
assets
s3 · hel-1
Runs on the servers and tools you already use
- Hetzner
- OVH
- Latitude.sh
- Bare metal
- On-prem
- GitHub
- GitLab
- Docker
- Cloudflare
- Slack
- Telegram
The cloud got expensive. Running your own servers got complicated.
Cloud repatriation went mainstream — 21% of workloads are already back on private infrastructure (Flexera, 2025). But leaving the cloud usually means hand-assembling a dozen tools nobody wants to own.
Cloud cost explosion
Bare-metal and EU hosts run 2.5–3.5× cheaper than hyperscalers for steady-state compute. 37signals cut ~$2M/yr leaving AWS — then replaced a $1.5M/yr S3 bill with hardware costing under $200k/yr to run.
The ops complexity tax
Self-hosting today means Terraform, Docker, a reverse proxy, TLS, Prometheus, Loki, backups, WireGuard, MinIO and k3s — each with its own failure mode. A dedicated SRE costs €90–120k/yr.
AI inference sticker shock
Managed inference is priced per token or per GPU-hour — an H100 endpoint runs ~$4,670/mo. The same open model on a rented GPU costs a fraction, if someone wires up vLLM, drivers, keys and metering.
Everything a modern cloud gives you — on hardware you control
Eight production primitives, one control plane, one price. Disaster recovery, GPU/LLM serving and Kubernetes are included — never add-ons.
Server onboarding
Bring any Linux server over SSH — or provision from your own Hetzner/OVH account. One-line agent install, deploy-ready in minutes, even behind NAT.
App deploys from Git
Push to GitHub or GitLab and ship. Dockerfile & compose detection, zero-downtime rollouts, automatic Let's Encrypt TLS and preview environments.
Managed databases
PostgreSQL, MySQL, Redis and MongoDB with verified, restorable backups to any S3-compatible target. Point-in-time recovery on the roadmap.
S3 object storage
Turn a storage box into S3-compatible buckets — MinIO or SeaweedFS — with scoped access keys, quotas, versioning and usage metering.
GPU / LLM inference
Serve open models on your own GPUs. OpenAI-compatible endpoints, managed API keys, per-key token metering and DCGM GPU telemetry — drivers bootstrapped for you.
Monitoring & alerting
Metrics, logs and alerts for every server and resource, zero-config. Sane default alerts on creation — alerting is opt-out, not opt-in.
Disaster recovery
Per-environment recovery plans with continuously replicated backups, health-checked automatic failover and DNS flip. Target RTO < 15 min, RPO ≤ 5 min.
Kubernetes clusters
Group owned servers into a managed HA k3s cluster over the mesh, with one-click upgrades and revocable kubeconfig export. No distro maintenance.
Disaster recovery, GPU/LLM serving and Kubernetes included — never add-ons.
One control plane over your infrastructure
SigmaHub orchestrates typed servers over an encrypted WireGuard mesh. You keep the hardware; we keep it managed — and we never resell it.
SigmaHub control plane
Orchestration · scheduling · metering · observability
General
apps · k8s
Storage
S3 · backups
Database
postgres · redis
GPU
LLM · inference
Servers you own — connected over an encrypted mesh, orchestrated from one control plane. SigmaHub never resells hardware.
Secure by default
WireGuard mesh between control plane and every server, outbound-only signed agents, and envelope-encrypted secrets (AES-256-GCM).
Works behind NAT
Agents dial out only — no inbound ports beyond your app's 80/443. Onboard servers with no public SSH over an optional VPN or jump host.
Full audit trail
Every key, deploy and failover is logged. Signed, cosign-verified agent releases and a SOC 2 / GDPR compliance roadmap.
Coolify's economics. A GPU cloud's inference. Railway's UX. Unified.
The deploy-PaaS tools solved one slice. SigmaHub is the whole platform — on servers you own, secure by default.
| Capability | SigmaHub | Coolify / Dokploy | Portainer / Rancher | Railway / Render / Fly | GPU clouds |
|---|---|---|---|---|---|
| App deploys on your own servers | partial | ||||
| First-class S3 storage on your servers | |||||
| GPU/LLM serving with keys + token metering | minimal | on their GPUs | |||
| Automated DR with failover + DNS flip | implicit | ||||
| Managed Kubernetes on owned servers | ops-heavy | ||||
| Security by default (mesh, signed agent) | basic | varies | managed | managed | |
| One all-inclusive price, never the servers | hobby tier | per-node | resold compute | GPU margin |
Own-server cost profile, cloud-grade experience
Keep the cost of hardware you control — pay a flat platform fee instead of metered public-cloud markups. Indicative, steady-state figures.
Single-GPU inference box, 24/7
H100 dedicated endpoint, 24/7
18 PB object storage
A 10-server fleet runs €50/month on SigmaHub — GPU, storage, DR and Kubernetes included — versus per-service, per-GB and per-token billing on public cloud.
Built for the teams the cloud priced out
From a solo agency running client apps to an AI team serving its own models — one mental model covers a single VPS and a 50-server fleet alike.
Agencies & indie devs
- Today
- Backup discipline per client, TLS toil, no per-client isolation.
- With SigmaHub
- One project per client, isolated environments, backups you can actually restore — starting free on up to 3 servers.
Startup CTOs
- Today
- PaaS bills that scale with traffic, vendor lock-in, no EU data story.
- With SigmaHub
- Railway-grade ergonomics on bare metal — typically 60–80% cheaper compute, with a platform fee that's a rounding error.
AI teams
- Today
- CUDA/driver toil, no token metering, data leaving the perimeter.
- With SigmaHub
- Self-hosted OpenAI-compatible inference on GPUs you control, with metering and keys — the wedge no rival serves.
SME platform teams
- Today
- ISO 27001 / GDPR audits, SSO, DR that's a binder not a system.
- With SigmaHub
- Audit logging, RBAC, encrypted secrets and rehearsed DR on infrastructure you fully control — SSO and fleet-scale orchestration as you grow.
From bare server to production in minutes
Install the CLI, connect a server, and deploy. SigmaHub handles the mesh, the runtime and the rollout — no SSH config, no Docker installs.
$ curl -fsSL https://get.sigmahub.io | sh$ sigma login$ sigma servers connect edge-fra-1$ sigma deploy ./webshop --env production✓ webshop deployed · https://webshop.acme.sigma.appOne number. The count of your connected servers.
That's the entire price list. Every feature — deploys, databases, S3, GPU/LLM, monitoring, disaster recovery and Kubernetes — is included. No add-ons, no egress fees, no per-token GPU billing, no per-seat tax.
Free
For side projects and your first deploys.
Start free- Up to 3 connected servers
- Every feature included — nothing gated
- Git deploys, databases & S3 storage
- GPU/LLM serving & Kubernetes
- Monitoring with default alerts
- Community support
Cloud
Scale past three servers. Same everything, one meter.
Start free- Unlimited connected servers
- Disaster recovery & automatic failover
- GPU/LLM serving with token metering
- Managed Kubernetes (k3s HA)
- Verified backups & audit log
- No add-ons, no egress, no surprises
Your first three connected servers are always free. Connect a fourth and every server bills at €5/mo — SigmaHub never sells, rents or marks up your infrastructure. That's the whole deal.
Questions, answered
Bring your servers. We'll bring the cloud.
Connect your first three servers free and deploy in minutes.