Skip to content
pipe pipe

Introduction

Pipe is a minimal, dependency-free pipeline runner for the command line. It replaces fragile, chained shell aliases with clear, structured YAML workflows that are easy to maintain, share, and collaborate on.

Why Pipe?

Shell scripts grow messy. A deploy command starts as docker build && docker push && ssh ... and quickly becomes an unmaintainable one-liner. Pipe gives that one-liner a proper home:

name: deploy
steps:
  - id: build
    run: "docker build -t myapp ."
  - id: push
    run: "docker push myapp"
    depends_on: "build"
  - id: deploy
    run: "ssh prod 'docker pull myapp && docker restart myapp'"
    depends_on: "push"
    retry: 2

Run it with pipe deploy.

Capabilities

  • Sequential and parallel execution — steps run in order by default, or in parallel when you use a list of commands
  • Dependency graph — explicit depends_on declarations and implicit $PIPE_* variable edges, with cycle detection
  • Output passing — each step’s stdout is captured as PIPE_<STEP_ID> for downstream steps
  • Named sub-runs — parallel commands with individual output capture
  • Variable templating — Go text/template syntax with three-layer precedence (YAML, environment, CLI)
  • Step caching — skip expensive steps that haven’t changed, with duration or wall-clock expiry
  • Sensitive data — redact outputs from state files, always re-execute on resume
  • Resume failed runs — pick up where you left off with --resume
  • Retry on failure — configurable per-step retry count
  • Hub (Beta) — push, pull, and version pipelines with a central registry

Next steps