# How It Works (/docs/how-it-works)

The simple mental model behind Solaris AI Flow: build, trigger, execute.



Every automation in Solaris AI Flow follows three steps:

1\. Build [#1-build]

Open the visual editor. Drag nodes onto the canvas: triggers, AI models, DeFi protocols, messaging bots, and utility steps. Connect them with edges to define data flow.

Each node takes input from connected upstream nodes and passes its result downstream.

2\. Trigger [#2-trigger]

Choose what starts your workflow:

* **Manual** - click Run in the editor
* **Webhook** - an external service POSTs to your unique URL
* **Cron** - runs on a schedule (every 5 minutes, hourly, daily, weekly)

3\. Execute [#3-execute]

When triggered, the runtime follows your graph in dependency order:

```
Trigger → Node A → Node B → Node C → Output
```

Each node executes as its dependencies become available. If Node B needs data from Node A, it reads Node A's output via template expressions like `{nodeAResponse.result.price}`.

Builder vs. Runtime [#builder-vs-runtime]

The **canvas** is where you design. It's a visual editor: drag, connect, configure.

The **runtime** is what runs. It reads your saved workflow, resolves credentials, calls external APIs, and records every node's output and errors for debugging.

You design on the canvas. The runtime executes what you designed.

Next steps [#next-steps]

* [Quickstart](/docs/getting-started/quickstart) - build your first workflow
* [Core Concepts](/docs/concepts) - understand workflows, nodes, and edges