Key concepts

This is the "how does this actually work?" section. Not a wall of jargon. The mental models you need so the assistant on your Linux machine is usable, not mysterious.

You do not have to read these in order.

• The workspace : the assistant's brain is a folder of plain-text files on disk.
• Skills and tools : tools are atomic actions. Skills bundle tools plus instructions.
• Memory and context : persistent knowledge versus what fits in the current turn.
• Model profiles : which LLM runs which job. You bring the keys.
• Channels : desktop app, Telegram, Slack, phone, and other ways to talk.
• Scheduling : one-shot and recurring work on the assistant's clock.

Words like hatch, guardian, home, and assistant are defined in the glossary. What the assistant can do out of the box is in Skills reference. Approvals live in Permissions.