01 · Start

Installation, in under a minute.

Reyn ships as a single Python package with no service to run. Install it, define a flow, and start orchestrating agents from your terminal.

Requirements

Reyn supports Python 3.10 and later on Linux, macOS, and Windows. You'll also need an API key for the LLM provider you plan to use — OPENAI_API_KEY, ANTHROPIC_API_KEY, or any provider supported by the runtime.

Install with pip

The fastest path is pip. Reyn has a small dependency footprint and installs in seconds.

terminal

$ pip install reyn

Verify the install with the version command:

terminal

$ reyn --version
# reyn 0.4.2

Note

If you're on a managed Python environment, use pipx install reyn instead so the CLI lands on your PATH without polluting the system interpreter.

Your first flow

A flow is a graph of phases. Each phase invokes one or more agents, and the runtime decides — based on LLM output — which phase to enter next.

flow.py

from reyn import Flow, Agent

researcher = Agent(
    name="researcher",
    skill="web_search",
)

writer = Agent(
    name="writer",
    skill="draft_summary",
)

flow = Flow("summarize")
flow.add(researcher).then(writer)

if __name__ == "__main__":
    flow.run(input="open-source AI orchestrators 2026")

Run it:

terminal

$ python flow.py

What just happened

The researcher agent ran first, called the web_search skill, and produced a structured result.
The runtime passed that result to the writer agent.
Every step was logged to ~/.reyn/runs/ — replayable and inspectable.

Tip

Pass --steer to reyn run to pause between phases and approve, edit, or reject what the next agent sees.

← Previous

Introduction

Your first flow