Close the gap between local dev and live K8s

What it does

mirrord runs a local process—whether launched by you or an AI agent—and makes it behave like a pod inside a live Kubernetes cluster. It routes outgoing traffic, file reads/writes, and environment variable lookups through a target pod you select, while mirroring that pod’s inbound traffic back to your machine. You get end-to-end validation against real services and data without pushing a build or disturbing the cluster.

The interesting bit

The tool launches a small agent pod on the same node as your target to act as a two-way bridge, which is more invasive—and more complete—than a simple network proxy. The README also emphasizes first-class support for AI coding agents like Claude Code and Cursor, letting them run generated code against live cluster context without ever touching a deployment manifest.

Key highlights

• Available as a VS Code extension, IntelliJ plugin, or CLI tool.
• Mirrors incoming traffic from the target pod to your local process and routes outbound calls through the cluster.
• Shares the target pod’s filesystem and environment variables with your local code.
• Supports AI agents including Claude Code, Cursor, Codex CLI, and Gemini CLI.
• Used by monday.com, SurveyMonkey, Cadence, and others.

Caveats

• The agent pod requires elevated Linux capabilities (CAP_NET_ADMIN, CAP_SYS_PTRACE, CAP_SYS_ADMIN, etc.), and stripping them via configuration can break functionality or render mirrord unusable in some environments.

Verdict

Grab it if you develop microservices that are hard to test outside a live Kubernetes environment and you want to skip the deploy-wait loop. If you don’t use Kubernetes, or your stack is already trivially reproducible locally, you don’t need this.