Stepflow Introduction
Stepflow allows you to create and execute AI workflows combining components from different tools and services, both locally and in the cloud. With Stepflow, components may execute locally or remotely, allowing simple development while providing isolation and resource management for production scale.
Stepflow defines a protocol for component servers, allowing a combination of custom and off-the-shelf components to be combined within a single workflow. By routing specific component servers to different Stepflow runtimes, you can create workflows that run across multiple machines, containers, or cloud services. Its modular architecture ensures secure, isolated execution of components—whether running locally or deployed to production.
Stepflow further solves for production problems like durability and fault-tolerance by journalling the results of each component execution, allowing workflows to be resumed from the last successful step in the event of a failure without adding complexity to component servers.
Architecture
Stepflow consists of a runtime that manages the execution of workflows and servers that provide components and tools using the Stepflow protocol or Model Context Protocol.
- Local
- Production
During development, the Stepflow runtime can manage the component servers and MCP servers in subprocesses, communicating over stdio.
In production, the Stepflow runtime can communicate with remote servers in separate containers or k8s nodes. This allows sharing a server across multiple runtimes and isolating specific components on dedicated servers for better security and resource management.
Next Steps
- Get Started by installing Stepflow and running your first flow.
- Read more about writing your own Workflows.
- Learn about available components and creating your own in Components.