Skip to main content

Documentation Index

Fetch the complete documentation index at: https://docs.sirenspec.dev/llms.txt

Use this file to discover all available pages before exploring further.

​
What is SirenSpec?

SirenSpec lets you define multi-agent AI workflows in human-readable YAML and execute them against OpenAI, Anthropic, or Ollama backends. You describe what each agent does and how agents connect — SirenSpec handles the rest. 
version: "0.1"

agents:
  assistant:
    model: "openai:gpt-4o-mini"
    system: "You are a helpful assistant. Answer questions clearly and concisely."

nodes:
  answer:
    agent: assistant
    writes: output.reply

input:
  message: "What is the capital of France?"

sirenspec run workflow.yaml
# → { "output": { "reply": "Paris" }, ... }

​
Key Features

YAML Workflows

Define agents, nodes, edges, and guardrails in a single readable YAML file.

Multiple Providers

Works with OpenAI, Anthropic, and Ollama out of the box.

Guardrails

Built-in injection detection and output length limiting on every agent call.

CLI + SDK

Run workflows from the command line or import directly into Python code.

​
How It Works

  1. Define agents — each agent has a model URI (provider:model) and a system prompt.
  2. Define nodes — each node binds an agent to an output path in the workflow context.
  3. Connect nodes with edges — edges control execution order and support conditional branching via when: expressions.
  4. Run — SirenSpec executes nodes in topological order, routing outputs through the context, and returns a structured JSON trace.

​
Next Steps

Quickstart

Install SirenSpec and run your first workflow in minutes.

YAML Reference

Complete reference for all workflow fields and options.