CP-02 · Service
Pull-through cache for every container registry
Container Mirror caches images from public and private registries in the region of your runners and clusters. Pulls get faster, rate limits stop applying to images you already use, and digests stay identical.
- Registries
- Docker Hub, GHCR, Quay, registry.k8s.io, MCR, ECR Public
- Endpoint
- acme.cacheport.cc/<registry>/<image>
- Clients
- Docker, Podman, containerd, CRI-O, BuildKit
- Tags
- Re-checked every 5 minutes
- Layers
- Cached by digest, deduplicated
- Artifacts
- Images, Helm charts and other OCI artifacts
What you get
Everything in Container Mirror
One host, many registries
Each upstream gets a short prefix: /hub, /ghcr, /quay, /k8s, /mcr. Add private registries with their own credentials.
Identical digests
Manifests and layers are served byte for byte, so image@sha256 references and signatures keep verifying.
Rate limit relief
Cacheport pulls upstream with one authenticated account. Your runners and nodes never hit the upstream directly.
Layer deduplication
Layers shared between images are stored once, so caching many tags of the same base image costs little.
Outage protection
If an upstream is down, cached tags keep resolving to their last known manifest.
Kubernetes ready
containerd and CRI-O can route docker.io and other registries through the mirror without changing image references.
Pull audit
Every pull is logged with token, image, digest and outcome.
Multi-arch
Each platform in a manifest list is cached as it is pulled.
Start
Up and running in three steps
Log in
Create a read token and log in from your CI or node.
terminal $ echo "$CACHEPORT_TOKEN" | docker login acme.cacheport.cc -u ci --password-stdin Login SucceededPrefix your images
Add the registry prefix to image references.
config node:22-alpine -> acme.cacheport.cc/hub/library/node:22-alpine ghcr.io/org/app:1.4 -> acme.cacheport.cc/ghcr/org/app:1.4Or mirror transparently
On containerd, route docker.io through Cacheport without touching manifests.
ci log # /etc/containerd/certs.d/docker.io/hosts.toml server = "https://registry-1.docker.io" [host."https://acme.cacheport.cc/v2/hub"] capabilities = ["pull", "resolve"] override_path = true
| Supported upstreams | Docker Hub, GitHub Container Registry, Quay.io, registry.k8s.io, Microsoft Container Registry, Amazon ECR Public, any OCI-compliant private registry |
|---|---|
| Regions | Frankfurt (fra1), Ashburn (iad1), San Jose (sjc1), Singapore (sin1), Sydney (syd1) |
| Protocol | OCI Distribution Specification, Docker Registry HTTP API V2 |
| Tag freshness | 5 minutes by default, configurable per upstream |
| Largest layer | 20 GB |
| Authentication | Token via docker login, OIDC from GitHub Actions and GitLab CI, IP allowlists for nodes |
Ecosystems
Setup guides by package manager
Common uses
Where it pays off first
Shared CI runners
Base images come from the cache instead of counting against Docker Hub limits for a shared address.
Kubernetes clusters
Node scale-ups pull images from the same region in seconds.
Air-gapped promotion
Pin the images you approve and serve them to environments that cannot reach the internet.
FAQ
Questions
Do I have to change every image reference?
Not on containerd or CRI-O, where a hosts configuration routes docker.io through Cacheport. With Docker, prefix the references or use a build argument for the registry.
Can I push images?
No. Container Mirror is pull-only. Push to your own registry and add it as an upstream if you want it cached.
Are signatures preserved?
Yes. Cosign signatures and attestations are OCI artifacts and are mirrored like images.
How long are layers kept?
As long as a cached manifest references them, within your plan's storage. Unused layers are evicted least-recently-used.
Works with the rest of Cacheport
Same workspace, same tokens
Start on the free plan
20 GB of cache, every registry, unlimited users. Upgrade when your pipelines outgrow it.