Cursus¶
Cursus turns a pipeline graph plus a JSON configuration into a complete, production-ready Amazon SageMaker pipeline — resolving inter-step dependencies, wiring inputs and outputs, and generating the SageMaker step objects automatically. You describe what the pipeline is; Cursus figures out how to build it.
Graph-to-pipeline automation — a DAG of step names compiles to a full SageMaker pipeline.
Intelligent dependency resolution — step connections and data flow are inferred, not hand-wired.
Declarative, data-driven steps — every step is one
<step>.step.yamlinterface; builders are synthesized at runtime.A pre-built pipeline catalog — 44 validated DAGs across 8 frameworks.
Agent-ready — a framework-neutral 70-tool MCP surface mirrors the CLI/API for LLM agents.
pip install cursus
cursus compile -d my_dag.json -c my_config.json -o pipeline.json
New here? Start with Getting Started.
Getting Started
Tutorials
Concepts & Guides
- Concepts, Architecture & Design
- System Architecture
- Design Principles
- DAGs & Compilation
- Dependency Resolution
- Step Interfaces & Builder Synthesis
- The Configuration System
- Registry & Interface-First Discovery
- Step Packs: External Steps
- Validation & Alignment
- The Pipeline Catalog
- Path Resolution & the Caller Hook
- The Agent (MCP) Tool Surface
- Execution Documents
- How-to Guides
- Migration
Reference