Documentation

Getting started, concepts and API reference.

1. Connect an AI model

Go to Integrations → AI models and add a provider. AgentOS speaks three protocols: OpenAI-compatible (OpenAI, Azure OpenAI, Mistral, Groq, DeepSeek, OpenRouter, Together, Ollama, vLLM, LM Studio…), Anthropic and Google Gemini. Enter prices per million tokens to enable cost tracking. A sandbox model is included for trying things out offline.

2. Add agents

Pick a template from the Agent Marketplace or describe the agent in plain English. Each agent has instructions, an operating mode, an output schema and optional knowledge bases. Every change creates a new version, and you can roll back.

3. Build a workflow

In Workflow Studio, drag a trigger, agents and actions onto the canvas and connect them. Reference earlier results with templates such as {{ nodes.sales.result.total }} or {{ trigger.email }}. Sending email or calling an API requires an approval step linked to that action.

4. Test, publish, run

Test run executes the draft with sandbox delivery. Publish creates an immutable version. Running executions always keep the version they started with. Webhook triggers receive a signed URL: sign the raw body with HMAC-SHA256 and send it as X-AgentOS-Signature.

API

A public API reference will be published here in a later release.