Welcome

Short videos that teach you to build agents on Mindset and run them in production.

Season 1 · Episode 1

How Mindset fits together

Build in your AI agent, run it in Mindset: the two halves, the pieces you build with, and the four things only a person does Read the docs

Transcript

Season one, episode one. How Mindset fits together.

In this video: the two halves of Mindset, the pieces you'll build with, and the four things only a person can do.

Mindset works in two halves. You build in your AI agent: you describe what you want, and it builds it through the Mindset MCP. Then it runs in Mindset, where your people and your systems use it. Any AI agent that can connect to an MCP server works, like Claude Code, Codex or Copilot. We'll use Claude Code in this series.

Everything you build sits in an environment. Every org starts with Sandbox, a safe place to build and try things out. You can add others, like Production, when you're ready.

An agent is the thing that acts. It has a purpose, a personality, background knowledge, rules it must follow, and a model.

For most work, a well-written agent is all you need.

A connection is the only way data gets in or out. It points at one system, holds its key, and offers named operations, like list courses or enrol a learner. An agent is only ever given the operations it needs, never the whole system.

Two more pieces, used only when you need them.

A script is for steps the agent must not skip, like waiting for a manager's answer.

A function does the same work exactly the same way, once per row.

Widgets show forms and your own layouts in a conversation. Skills are instructions you write once and share across agents, like a house style.

Work reaches an agent in four ways: someone chats with it, a schedule wakes it, your own server calls it, or another system's webhook starts it.

Every time your AI agent saves a change, it's a new version, and nothing runs until a person activates it. Going back is just activating an earlier version.

And four things are always left to a person, never to your AI agent. Activating a version. Switching things on, like a new agent or connection. Entering keys and passwords, through a secure link the agent never sees. And inviting people to your org.

Next, we'll connect Claude to Mindset.

Playlists

Episodes in an order that makes sense. Press play and they run one after another.

Seasons

Each season builds on the last. New to Mindset? Start with Season 1.

3 playlists

Learning paths

Hand-picked episodes from across the seasons, around one theme.

2 playlists

All episodes

Every episode, by season.

Season 1

Build your first agent

From connecting Claude to switching your first agent on

5 episodes · 12 min

Season 2

Ship it like software

Versions in your repo, tests, Production, triggers and embedding

5 episodes · 4 min

Season 3

Make it do more

Scripts, approvals, functions, widgets, messages and insight

9 episodes · coming soon