Terminal
helm install mirrord-operator metalbear/mirrord-operator
Install the mirrord Operator
One Helm install on your existing staging cluster. The Operator handles per-PR isolation, traffic routing, queue splitting, and database branching.
mirrord Preview Environments
Instantly Preview Your Pull Requests in a Real Environment
mirrord gives every pull request an isolated preview environment on your shared staging cluster, connected to real services, real databases, real queues.
No ephemeral clusters
Seconds, not minutes
Request-level isolation
Trusted by engineering teams at
Ticket → preview → feedback
Open a PR, get a live preview.
Every PR — opened by a developer or by an AI agent — gets a live preview URL the moment it opens. Reviewers click the link and see the change running against real services, no deploy required.
Get started
Up and running in minutes
Three steps to set up preview environments on your existing staging cluster. No new infrastructure to provision. Read the full setup guide →
.github/workflows/preview.yml
- uses: metalbear-co/mirrord-preview@main
with:
action: start
target: deployment/my-app
namespace: staging
image: myrepo/myapp:${{ github.sha }}
key: pr-${{ github.event.pull_request.number }}
with:
action: start
target: deployment/my-app
namespace: staging
image: myrepo/myapp:${{ github.sha }}
key: pr-${{ github.event.pull_request.number }}
Wire it into your CI
Add one step to your CI workflow. On every PR, mirrord deploys the WIP image and gives that PR its own isolated session.
action output
✔ Header: baggage: mirrord-session=pr-142
Open a PR, see it live
The action outputs a unique session header per PR. Hit your staging URL with that header and you see the change live against real services — devs, PMs, designers, QA all share the same preview. Close the PR and the preview tears itself down.
Not on the engineering team?
Forward this page — one Helm install plus one CI step is all your platform team needs.
Capabilities
Everything you need for production-grade previews
No custom tooling. No ephemeral clusters. Just real previews on your real cluster.
🔒
Request isolation
The mirrord Operator filters traffic so only tagged requests reach your preview deployment. Other developers and services are unaffected.
📨
Queue splitting
Preview environments can consume from real message queues (Kafka, SQS, RabbitMQ) with automatic splitting — no duplicate processing.
🗄️
Database branching
Each preview gets an isolated copy-on-write branch of your real database. Test schema changes and seed data against production-shaped data, with no risk of polluting shared state.
☁️
Zero per-PR infrastructure
No clusters to provision, no namespaces to manage, no Terraform to maintain. Preview deployments run on your existing shared staging cluster.
🔄
CI/CD integration
Works with GitHub Actions, GitLab CI, CircleCI, Jenkins, and any pipeline that can run a CLI command. Drop-in setup, no vendor lock-in.
🌐
Any language or framework
mirrord works at the syscall level. Rust, Go, Node.js, Python, Java, .NET, Ruby — if it runs on Linux, it works with mirrord.
Who it's for
Every role in the PR loop
Per-PR previews run on your existing shared staging cluster. The mirrord Operator handles isolation, RBAC, and traffic routing.
- — Zero per-PR infrastructure
- — Works with your existing cluster
- — Operator-level isolation, no leaks
Your cluster
staging-shared●
↳ pr-1247●
↳ pr-1248●
↳ pr-1252●
Every PR is reviewed against real services and real data. Reviewers see what production will see — before merge.
- — Real integration testing per PR
- — Faster cycle from open to merge
- — Fewer staging escapes, fewer rollbacks
PR #1248 · Checks
Unit testspassed
Lintpassed
Preview vs real serviceslive ↗
Integration suitepassed
Click any open PR and get a live preview against real data. Review the actual experience, leave feedback, ship.
- — Live preview link per PR
- — Real data, no mocks
- — Feedback before merge, not after
PR #1248 · Preview
★ New feature live here
FAQ
Got a question?
How is this different from spinning up an ephemeral cluster per PR?
Ephemeral clusters provision a full copy of your environment for every PR — expensive, slow, and prone to drift. mirrord preview environments run on your existing shared staging cluster. The mirrord Operator isolates traffic at the request level, so you get the same isolation guarantees without any per-PR infrastructure.
Do preview environments work with message queues?
Yes. mirrord supports queue splitting for Kafka, SQS, and RabbitMQ. Messages destined for the service under preview are automatically routed to the preview deployment, while other messages continue to flow normally.
Can multiple PRs be previewed at the same time?
Yes. Each preview session is isolated by the mirrord Operator. Multiple developers can have concurrent preview environments on the same cluster without interfering with each other.
What CI/CD systems does this work with?
mirrord works with any CI/CD system that can run a CLI command — GitHub Actions, GitLab CI, CircleCI, Jenkins, and more. See our docs for setup guides.
Which plan includes preview environments?
Preview environments are available on the Enterprise plan. See pricing for details, or book a demo to get set up.
Get started
Every PR deserves a real preview.
Works with your existing staging cluster. Live in minutes.
Read the docs →