cacheport

Package & build cache for CI

Builds that don’t wait on the internet.

Cacheport is a caching proxy for npm, PyPI, Maven, Go, Cargo, Docker Hub and 12 more registries, plus a remote build cache. Point CI at one URL and every dependency after the first comes from a cache next to your runners.

Free plan: 20 GB of cache, every registry, unlimited users. No card needed.

Pull log · acme · fra1Live sample
  1. 14:02:07react@19.1.0 npm · gha-runner-07CACHE3 ms
  2. 14:02:06numpy-2.1.3-cp312 pypi · gitlab-r-12CACHE9 ms
  3. 14:02:06node:22-alpine hub · bk-agent-31CACHE41 ms
  4. 14:02:05golang.org/x/net@v0.31.0 go · gha-runner-02UPSTREAM212 ms
  5. 14:02:04serde-1.0.215.crate cargo · gha-runner-19CACHE2 ms
  6. 14:02:04spring-core-6.2.0.jar maven · jenkins-a4CACHE6 ms
  7. 14:02:03left-pad@9.9.9 npm · gha-runner-07COOLDOWN1 ms
  8. 14:02:02django-5.1.4.tar.gz pypi · gitlab-r-03CACHE5 ms
  9. 14:02:01esbuild@0.24.0 npm · gha-runner-11CACHE4 ms
96.4%
Hit ratio
18,240
Upstream calls saved
7 ms
Median hit

18 registries, one endpoint

Every package manager keeps its own client and lockfile. Only the address changes.

All ecosystems →

Also: Gradle, Go modules, Cargo, RubyGems, Composer, NuGet, Docker Hub, Helm, apt, Alpine apk.

Four caches, one workspace

Use one or all of them. They share tokens, regions, audit logs and a single bill.

From cold to cached in one pipeline run

  1. Step 1Workspace

    Create a workspace

    Sign up, pick the region closest to your runners and create a read token for CI.

    terminal
    $ cacheport login
    $ cacheport workspace create acme --region fra1
    $ cacheport token create ci --scope read
  2. Step 2Clients

    Point your clients at it

    Each package manager needs one setting. Store the token as a CI secret.

    config
    registry=https://acme.cacheport.cc/npm/
    index-url = https://acme.cacheport.cc/pypi/simple/
    GOPROXY=https://acme.cacheport.cc/go,direct
  3. Step 3Pipeline

    Run your pipeline

    The first run warms the cache. From the second run on, the pull log shows hits.

    ci log
    $ npm ci
    added 1214 packages in 9s
    # Cache-Status: cacheport; hit

Built for the dull failures that break builds

The registry is slow today

Cached packages come from your region in milliseconds, whatever the public registry is doing.

Docker Hub said “toomanyrequests”

Runners pull through one authenticated cache instead of hitting per-IP limits from a shared NAT address.

A package vanished upstream

Versions you have used stay in your cache, so old releases still rebuild.

A fresh release was malicious

Release cooldown keeps brand-new versions out of fresh installs until they have been public long enough to be vetted.

Nobody knows what a build pulled

Every pull is logged with repository, token, version and outcome, ready to export.

Cold caches on ephemeral runners

New runners start with a warm regional cache instead of an empty disk.

Works with the CI you already run

Tokens go in your CI secrets. GitHub Actions and GitLab CI can skip long-lived tokens entirely and sign in with OIDC.

Common questions

How is this different from the cache built into my CI?

A CI cache restores files on a runner that has seen the project before. Cacheport serves packages to any runner, from any repository in your workspace, and keeps working when the public registry does not.

Do I need to change lockfiles?

No. Clients keep verifying every artifact against your lockfile hashes. Only the registry address changes.

Where is my cache stored?

In the region you pick when you create the workspace: Frankfurt, Ashburn, San Jose, Singapore or Sydney. Business and Enterprise plans can add more regions.

What does it cost?

The Free plan is free forever. Team is $29 a month and Business is $149 a month. See pricing for limits.

Start on the free plan

20 GB of cache, every registry, unlimited users. Upgrade when your pipelines outgrow it.