10x Cheaper
Fraction of the cost of hyperscalers. $1/GB RAM/month licensing — predictable, transparent, no billing surprises. Ever.
10x Simpler
Do months of work in a few days. A clean UI to manage your entire cluster — deploy, scale, and monitor from one dashboard.
100% Yours
Self-host across any provider. No vendor lock-in, no data leaving your control. Switch providers at will without changing a single line of config.
Predictable Pricing. No Surprises.
Starting at $16/month for 16 GB of RAM and 20 users. Beyond that, licence cost $1 per GB of RAM per month and $5.99 per user per month. That's it. No egress fees, no NAT gateway costs, no monitoring surcharges, no data-transfer markups. Bolt-on any hardware you want, and Cosmos takes care of the rest.
* Waivable with custom pricing — contact sales.
AWS pricing based on us-east-1 on-demand m6i.4xlarge instances, March 2025. Hardware pricing based on Hetzner AX52 dedicated servers. Actual prices may vary by region and configuration.
Hours. Not Months.
Reverse proxy, HTTPS, SSO, VPN, WAF, monitoring, backups, load balancing — Cosmos ships all of it pre-wired. Building the same stack yourself means gluing together a dozen tools, each with its own config, upgrades, and failure modes.
Rough estimates for an engineer already familiar with each tool, targeting HTTPS, SSO, reverse proxy, monitoring, VPN, WAF, backups and failover across 3 servers. Ongoing maintenance not included.
Do You Really Need a Hyperscaler?
Probably not. Here's why.
The difference is about 1 hour per year. Your application code or database will hit its own limits long before the underlying infrastructure does.
It's your code, your database queries, your architecture. Paying 5x more for infrastructure doesn't fix application-level problems.
Uptime comes from good architecture, not expensive providers. A well-configured cluster on dedicated hardware achieves the same real-world reliability.
If you're not processing millions of transactions per second, you're paying for capacity you'll never touch.
The Per-Hour Pricing Trap
Let's do the math.
Run it 24/7 and that's $560/month per server. Per-hour pricing exploits the gap between how the price feels and what it actually costs.
The only workloads where per-hour pricing makes sense are batch jobs on idle machines — and those are cheap anyway. Your production servers never sleep, so per-hour is just monthly pricing dressed up to look smaller.
A properly sized flat-rate cluster handles traffic bursts gracefully, without a surprise bill for auto-scaled capacity you didn't ask for.
$1/GB RAM/month. No calculator needed, no bill shock at end of month.
Everything You Need to Scale
Cosmos Pro is the exact same software as Cosmos Cloud, with the building blocks of a real platform on top: multi-server deployments, managed databases, S3 storage and fine-grained permissions.
Run one app on many servers
Describe the app once, say how many copies you want, and Cosmos does the rest.
Your Cosmos servers connected through Constellation already form a cluster. A Deployment is the same compose file you use for a ServApp, run as replicas across your nodes and served behind a single URL.
- Fixed count, Autoscale (min / max), or Fill: one replica on every node that matches your tags
- Placement by tags (gpu, nvme, zone-eu) and by load: round-robin or least-busy
- Rolling updates one node at a time. A failed replica pauses the rollout, the rest keep serving
- A node goes down: replicas are rescheduled within seconds. A leader goes down: another manager takes over
- Scale to zero: idle tools stop after an hour and restart on the next request
PostgreSQL in one click
Private to your network, one database per app, backed up every night.
Create a PostgreSQL server from the dashboard, then hand each application its own database with its own user and password. Nothing is reachable from outside your Constellation unless you say so.
- Per-application databases with isolated credentials and one-click password rotation
- Reference credentials from any compose with template variables, so secrets never live in your config
- Scheduled restic backups to disk or any remote storage, with retention policies
- Restore any snapshot as a new database next to the live one. The original is never touched
- Size, connections and resource graphs, plus a full event history
DATABASE_URL=${db.maindb.wiki.url}Your own S3, spread across your servers
Anything that speaks S3 can use it. Every file is copied to another node.
Tag the nodes that should hold data and Cosmos deploys an S3-compatible storage across them, powered by SeaweedFS. Backup tools, media apps, Nextcloud, your own code: point them at the endpoint and go.
- Choose your replication level: survive losing one node, or two
- Grow by tagging another node. Shrink by draining one before you remove it
- Automatic maintenance: vacuum, replica scrubbing, health checks, optional erasure coding for cold data
- Highly available S3 gateways behind your load balancer, with valid HTTPS on your own domain
- Private to Constellation by default. Make it public in one click when a remote service needs it
Exactly the access each person needs
Go beyond User and Admin with groups you define yourself.
A group is a name and a set of permissions. Put someone in it and they get precisely those rights: a helpdesk that can invite users but never touch the server, operators who can restart apps but not change URLs, auditors who see everything and change nothing.
- Nine granular permissions split between read and manage: logs, users, containers and storage, configuration, secrets
- The menu adapts to the group. Pages a member cannot use are simply not there
- sudo-style privilege elevation: sensitive actions stay locked until re-authenticated, for admins too
- Groups are shared across every server in your Constellation
- Scoped API tokens with IP whitelist, network restriction and expiry for scripts and Terraform
And everything around it
Plus all of Cosmos Cloud: reverse proxy, HTTPS, SSO, app market, backups and the Constellation VPN.
Built-in Load Balancer
One URL for all replicas, wherever they run. Round-robin or nearest-healthy, sticky sessions, and unhealthy replicas pulled from rotation within seconds.
SmartShield Enterprise
Adaptive rate limiting, anti-DDoS, bot protection and geo-blocking on every route, automatically. No WAF to configure.
Constellation Mesh
Encrypted network between all your servers, across providers and regions. Databases and storage stay private to it by default.
Monitoring & Events
CPU, memory, network and traffic graphs for every deployment, database and storage, plus a history of everything the cluster did.
Scheduled Jobs
Backups, maintenance and your own tasks run from one scheduler, on the node that owns the data, with retention policies.
Functions & Registries
Cloud Functions and private Package Registries backed by your object storage. In beta today, included in Pro.
Your Global Infrastructure
Deploy clusters across regions and providers. Every connection encrypted, every request load-balanced, every server under your control.
All inter-cluster traffic is encrypted end-to-end via Constellation VPN. Requests are automatically routed to the nearest healthy node.
How Cosmos Pro Compares
More features, less complexity, a fraction of the cost
| Feature | Cosmos Pro | AWS / GCP / Azure | DigitalOcean | Railway |
|---|---|---|---|---|
| Infrastructure | ||||
| Auto-Scaling Replicas | ||||
| Reverse Proxy + HTTPS | ||||
| Anti-DDoS / WAF | ||||
| Built-in VPN | ||||
| Security by Default * | ||||
| Ownership | ||||
| All Included in Base Price | ||||
| Predictable Pricing | ||||
| Self-Hostable | ||||
| No Vendor Lock-in | ||||
| Multi-Provider Support | ||||
* Security by Default: every service you deploy automatically gets HTTPS, authentication, rate limiting, DDoS protection, monitoring, and encryption — with zero configuration required.