SimpleAgentsEvery agentic SaaS is a config

Define your entire AI product as a YAML workflow. Run it in Python or TypeScript. Ship today.

Get Started in 5 Minutes

Examples

View on GitHub

Your product is a YAML

Classification, routing, extraction, generation -- describe your workflow as a graph config. No framework. No lock-in.

10 lines to production

pip install or npm install. Point at your YAML. Run. Stream. Done.

Multimodal out of the box

Text, images, streaming, structured JSON -- all from the same YAML config.

Observability included

Langfuse or Jaeger with one env block. OpenTelemetry native.

Self-healing JSON

LLM outputs truncated JSON? Auto-healed. Schema coercion built-in.

Rust-fast, everywhere

Rust engine. Python and TypeScript bindings. WASM for the browser. Same config runs everywhere.

The Idea

Every agentic SaaS -- email classifiers, document processors, intake systems, interview bots, support agents -- is fundamentally the same thing: an LLM workflow with structured outputs and deterministic routing.

SimpleAgents lets you define that workflow as a YAML config and run it with a single command. No framework to learn. No abstractions to fight. Just a graph of nodes and edges.

Choose Your Path

I want my first working workflow -> Workflow Quickstart
I want runnable examples -> Examples
I want Python API details -> Python Binding
I want TypeScript / Node API details -> Node.js Binding
I want to understand YAML routing and workers -> YAML Workflow System
I want tracing with Langfuse / Jaeger -> Tracing & Observability
I want to use Rust directly -> Rust Quick Start
I want to contribute -> Development Guide

Use It When

You want structured LLM workflows with deterministic routing
You want one YAML config that runs in Python or TypeScript
You want custom worker steps and tracing without building a framework around it

If you only need one prompt call and no workflow, start with the language bindings first.