# Ship your federation anywhere

Managed cloud or fully self-hosted. Helm charts for any Kubernetes cluster, Docker Compose for local development, Terraform for infrastructure as code, and live fleet monitoring—all built into Cosmo.

## Overview

## Every deployment model, built into Cosmo

Cosmo runs wherever your infrastructure runs. Use Cosmo Cloud for a fully managed experience, deploy to your own Kubernetes cluster with production-grade Helm charts, or spin up the complete stack locally with Docker Compose in minutes.

Once deployed, the Terraform provider manages your federation resources as code. Router Cluster Management gives you a real-time view of every running router instance. Router Compatibility Versions ensure that upgrades never silently break running deployments.

## Why deployment flexibility matters

## Why teams need real deployment choice

Most federation platforms offer one deployment model. Startups have no time to manage infrastructure. Regulated enterprises cannot send data to a third-party cloud. Platform teams need reproducible deployments and version-controlled changes. A single model cannot serve all three.

### Four deployment problems that come up repeatedly in production.

#### No time to manage federation infrastructure.
Setting up a control plane, schema registry, metrics pipeline, and CDN takes significant DevOps work. Teams building APIs should not also be running databases.

#### Compliance blocks cloud-hosted platforms.
Financial services, healthcare, and government organizations often cannot send API metadata to a third-party SaaS. They need all components running inside their own network.

#### Manual resource management causes configuration drift.
Clicking through a UI to create federated graphs and subgraphs does not scale across teams. Changes are hard to review, hard to roll back, and impossible to audit.

#### Router upgrades are risky without compatibility checks.
In a federated architecture, the router depends on execution configurations generated by composition. Breaking changes between them can cause silent failures that only surface in production.

Cosmo covers every model: managed cloud, full self-hosted, Docker for development, Terraform for automation, and built-in operational visibility.

## Cosmo Deployment capabilities

01 Managed service  
02 Self-hosted  
03 Automation  
04 Operations

## Which deployment capability do you need?
| If you are… | Start here |
| --- | --- |
| Evaluating Cosmo without committing to infrastructure | [Docker](/content/cosmo/deployment/docker/index.html) |
| Starting a new project and wanting managed infrastructure | [Cosmo Cloud](/content/cosmo/deployment/cosmo-cloud/index.html) |
| Deploying to a Kubernetes cluster in production | [Kubernetes (Helm)](/content/cosmo/deployment/kubernetes/index.html) |
| Requiring complete data sovereignty or air-gapped deployment | [Self-Hosted Deployment](/content/cosmo/deployment/self-hosted/index.html) |
| Managing federation infrastructure through version control | [Terraform Provider](/content/cosmo/deployment/terraform/index.html) |
| Monitoring CPU, memory, and version across deployed routers | [Router Cluster Management](/content/cosmo/deployment/cluster-management/index.html) |
| Upgrading routers safely without risking configuration mismatches | [Router Compatibility Versions](/content/cosmo/deployment/router-compatibility-versions/index.html) |

## How Cosmo deployment compares

|  | Cosmo | Cloud-only platforms | Custom self-hosted |
| --- | --- | --- | --- |
| Managed cloud option | Yes (Cosmo Cloud) | Yes | No |
| True self-hosted (full stack) | Yes (Helm chart) | No | Yes (build it yourself) |
| Production-grade Helm chart | Yes | No | Manual |
| Terraform provider | Yes (official) | Varies | Manual |
| Router fleet monitoring | Built-in | Varies | Manual |
| Free tier | Yes (10M req/month) | Varies | N/A (compute costs) |

## Deployment use cases

Real deployment patterns and the Cosmo capability behind each one.

### Startup velocity
#### Production federation in days, without a dedicated ops team
##### Scenario
A startup needs GraphQL federation but has no DevOps capacity to manage infrastructure. They need to ship quickly and stay within a tight budget.
##### How Cosmo handles it
Sign up for Cosmo Cloud, create federated graphs through Studio, and deploy routers to any environment. WunderGraph manages the control plane, Studio, and CDN. The free tier covers 10 million monthly requests.
##### Outcome
Production-ready federation running in days. Engineering time goes to building APIs, not managing infrastructure.

### Compliance & sovereignty
#### Federation that stays inside your network
##### Scenario
A financial services or healthcare organization needs GraphQL federation but cannot send data outside their private cloud. Regulatory requirements demand full data residency.
##### How Cosmo handles it
Deploy the complete Cosmo stack to a private Kubernetes cluster using the Helm umbrella chart. All components—Controlplane, Studio, metrics, CDN—run inside your network. Configure `values.yaml` with external databases for production durability.
##### Outcome
Full federation platform with no data leaving the organization. Regulatory requirements met without sacrificing federation capability.

### GitOps automation
#### Every infrastructure change reviewed, versioned, and audited
##### Scenario
A platform team wants federation infrastructure managed through pull requests, with previews of every change before it applies to production.
##### How Cosmo handles it
Store Terraform configurations in Git. Use the `wundergraph/cosmo` provider to define federated graphs, subgraphs, namespaces, and feature flags. Run `terraform plan` in CI for review; apply on merge.
##### Outcome
Complete audit trail of all federation infrastructure changes. Rollback is a git revert.

## Why teams use Cosmo for deployment
- **No lock-in.** Run Cosmo in your cloud, your Kubernetes cluster, or your own datacenter. Switch from Cosmo Cloud to self-hosted at any point. The same platform runs in every environment.
- **Production-grade tooling, not scripts you maintain yourself.** The Helm chart, Terraform provider, and Docker Compose configuration are maintained by WunderGraph and kept current with every platform release.
- **Operational visibility without extra instrumentation.** Router Cluster Management discovers running instances automatically via built-in OpenTelemetry. CPU, memory, version, and composition are visible in Studio the moment routers come online.

## Deploy your federation your way