# How Jarvis thinks
URL: https://jarvis.ceo/docs/concepts/how-jarvis-thinks
Status: beta

The mental model: a fast voice layer that answers and routes instantly, and a stronger agent that does the heavy work, in the foreground or quietly in the background.

> [beta] The agent behaviors here are in early access. Download the beta, or join the waitlist for full access.

Jarvis works the way a great assistant does: it answers the quick things instantly, and for the real work it makes a plan and does it. Knowing this makes it easier to ask for the right thing.

## Two layers

- A fast voice layer hears you, answers simple things, and routes the rest.
- A stronger agent handles multi-step work, either while you wait or in the background.

## It plans before it acts

For anything with more than one step, Jarvis writes a short numbered plan first, so you can see what it is about to do before it does it.

## It answers with the right shape

Rather than a wall of text, Jarvis draws the answer: a table to compare, a checklist for steps, a chart for trends. See generative UI cards.

## Foreground or background

Say this: "Look into this in the background and let me know when you have an answer."
Jarvis: On it. I will notify you when the result is ready.

> Checkpoint: You can hand Jarvis a goal and trust it to plan, act, and report. Next: how it gets better the more you use it.

Related: Multi-step chains (https://jarvis.ceo/docs/control/chains), Generative UI cards (https://jarvis.ceo/docs/pantry/card-types)