Tutorials

Build an AI agent, step by step

Each tutorial is a build you can follow along with: the video shows it happening, and the written steps are the reference you come back to. They are ordered - the first one is the whole loop at its simplest, and each one after adds a thing worth understanding on its own.

The tutorials

In order. Each one assumes the one before it, but none of them takes long.

What these have in common

Every build follows the same four moves, which is the real thing to take away: create the agent, give it a channel so people can reach it, give it tools so it can do something, then decide what it may do without asking. Once that shape is familiar, a connector you have never used before is a ten-minute job rather than a project.

The fourth move is the one worth slowing down for. An agent that can only read is useful and nearly risk-free; an agent that can write, send and change things is where the value is and where the care belongs. Trust levels let you move that line one action at a time, and every action is recorded either way - the reference is in approvals and trust.

If you would rather read than watch

The documentation covers the same ground as a manual rather than a walkthrough, and channels and connectors is the full reference for every channel and integration. The resources library is the why rather than the how - what agents are good at, where the boundaries belong, and how to choose a platform.

Follow along on your own deployment

Every tutorial here runs on a standard deployment - the archetype catalog, the connector picker and the approval gates are all in place from the start.

Start a deployment