Your agent workflows deserve better than scripts

Reuse pipelines across projects. Test before you ship. Collaborate on agent logic like you do code.

The problem every team faces

You write agent workflows in scripts or config files. Next project? You rewrite them. No testing framework. No way to share. No git history.

It's like writing infrastructure by hand every time.

PipelineAsCode changes that

Define once, reuse everywhere

Write agent DAGs in Python or TypeScript. Share them via our marketplace. Other teams fork, customize, and contribute back.

Test locally, deploy with confidence

Our CLI lets you run full DAGs end-to-end before production. Assert outputs. Catch bugs early.

Git-native workflows

DAGs as code mean readable diffs, full audit trails, and code review for your agent logic.

Primitives you actually need

Retry logic, gates, streaming, error handling—all built in. No boilerplate. No reinventing the wheel.

Real workflows, real teams

Teams building internal AI platforms are exploring how to standardize agent logic. We're learning from them and building templates like 'analyze-patch-test-pr' that work out of the box.

How we're thinking about this

The SDK is open source. You can test and deploy locally. We're offering optional cloud hosting for execution and monitoring—but you're never locked in.

Explore the docs

FAQ

Is this open source?

The SDK and CLI are open source. We're exploring a freemium model: cloud hosting and premium templates are paid; self-hosting is free.

Can I use this with my existing tools?

Yes. PipelineAsCode integrates with your existing agent libraries and LLM providers. It's a layer for orchestration and reuse, not a replacement.

What does the template marketplace look like?

Teams share DAG templates with versioning and tags. You can fork, customize, and contribute improvements. Think npm for agent workflows.

How do I get started?

Install the SDK, write a simple DAG, test it locally with our CLI, and push to git. Cloud hosting is optional.