Dagger is an open-source runtime for composable workflows. It's perfect for systems with many moving parts and a strong need for repeatability, modularity, observability and cross-platform support. This makes it a great choice for AI agents and CI/CD workflows.
-
Reproducible execution engine, powered by containerized functions and a declarative DAG scheduler.
-
Universal type system, for strongly typed composition and discovery, across platforms and languages.
-
Powerful data layer: out-of-the-box caching, immutable state, and data tracability.
-
Native SDKs for 5 languages. Go, Typescript, Python, PHP, Java - and more on the way.
-
Open ecosystem: Thousands of modules at your fingertips, all interoperable across languages and platforms.
-
Interactive command-line environment, for rapid prototyping and debugging.
-
Batteries-included observability. Deep tracing, metrics (including token count), and logs, all accessible from the CLI or a web UI.
-
Adapts to you. Seamlessly integrate with all major compute and storage platforms, CI systems, languages, and agent frameworks.
-
LLM augmentation. Connect to any LLM endpoint (OpenAI, Google, Anthropic, LLama, DeepSeek, etc.) and give it access to your Dagger objects. Dagger automatically handles the agentic loop. No complicated framework needed.
- Join the Dagger community server
- Follow us on Twitter
- Check out our community activities
- Read more in our documentation
Interested in contributing or building dagger from scratch? See CONTRIBUTING.md.