Dagger — CI/CD Pipelines as Code in Any Language

What it is

Dagger is a programmable CI/CD engine that lets you define pipelines in TypeScript, Python, or Go instead of vendor-specific YAML. Every pipeline step runs inside a container, so the behavior is identical whether you execute it on your laptop, in GitHub Actions, or in GitLab CI.

It targets platform engineers and DevOps teams tired of debugging YAML syntax errors, vendor lock-in, and the 'works on my machine' gap between local dev and CI environments.

How it saves time or tokens

Traditional CI systems require pushing code to trigger a pipeline, waiting for results, then iterating. Dagger pipelines run locally in seconds, giving you instant feedback. Because pipelines are real code, you get IDE autocomplete, type checking, unit tests, and code reuse — none of which YAML supports. Debugging shifts from reading CI logs to running a debugger.

How to use

1. Install the Dagger CLI with curl -fsSL https://dl.dagger.io/dagger/install.sh | sh or brew install dagger/tap/dagger.
2. Initialize a Dagger module in your project with dagger init --sdk=typescript (or python/go).
3. Write your pipeline functions, then run them with dagger call <function-name>.

Example

// dagger/src/index.ts
import { dag, Container, object, func } from '@dagger.io/dagger'

@object()
class MyPipeline {
  @func()
  async test(): Promise<string> {
    return dag
      .container()
      .from('node:20-slim')
      .withDirectory('/app', dag.host().directory('.'))
      .withWorkdir('/app')
      .withExec(['npm', 'install'])
      .withExec(['npm', 'test'])
      .stdout()
  }
}

Related on TokRepo

• DevOps tools — CI/CD, infrastructure, and deployment automation
• Automation tools — Workflow and task automation for developers

Common pitfalls

• Dagger requires Docker (or a compatible container runtime) running locally. If Docker Desktop is not running, dagger call fails immediately.
• Pipeline functions must be deterministic for caching to work. Non-deterministic steps (like fetching 'latest' tags) bust the cache on every run.
• The SDK learning curve exists even though you write 'real code.' Understanding Dagger's container chaining model takes a few iterations.