> ## Documentation Index
> Fetch the complete documentation index at: https://docs.moneypenny.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# Build with AI

> Describe what your agent should handle in your own words, watch it take shape, and test it — all on one screen.

The **Build** tab is the first tab in the agent editor and the fastest way to shape your agent. Tell Moneypenny AI what your agent should handle on a call, in plain English, and it builds the agent for you — the greeting, the journeys, the details to collect. You can fine-tune everything it builds on the other tabs whenever you like.

To get there, open **Agents** in the sidebar and pick your agent — the editor opens on the Build tab. Choosing **Build with AI** when you [create a new agent](/quickstart) lands you here too.

## The two halves of the screen

**Build with chat**, on the left, is where the building happens. It opens with: "Tell me what your agent should handle on a call and I'll build it. You can fine-tune everything else on the other tabs." Type what you want in your own words — the chat already knows your business from your business profile, so you don't need to explain who you are.

The **side panel** on the right shows you what you're building and lets you try it, without leaving the page. Switch between its three views with the buttons at the top:

* **Your agent** — a plain-English mirror of what's been built so far: your agent's details, what it says first, and what it can do — one card per journey. Select **Show steps** on a card to see its steps in order, with any branches as "If …" lines, and **Open in Journeys** to jump straight to that journey on the [canvas](/building/journeys). A card also names any connected apps the journey uses.
* **Chat** — a text conversation with your agent, so you can immediately try out what you've just described. Chat conversations use your draft, unpublished changes included.
* **Voice call** — a live voice call with your agent, right in the browser. A green dot appears on the button while a call is in progress. Voice calls always use the published version of your agent, so [publish](/building/publishing) first if you want to *hear* a change.

All three views keep their state when you switch between them — flipping over to **Your agent** mid-conversation doesn't lose your test chat or end your call.

## Build, watch, test

<Steps>
  <Step title="Describe what your agent should handle">
    Tell the chat what you want in a sentence or two — "Take messages when we're closed, and book table reservations", say. It may ask a couple of questions to fill in the gaps, then builds what you agreed.
  </Step>

  <Step title="Watch Your agent update">
    Each change appears in the **Your agent** view as it lands. Read it back as a caller would experience it — if something isn't right, just tell the chat and it will adjust.
  </Step>

  <Step title="Test as you go">
    Switch to **Chat** and have the conversation a real caller would. Ask for the things you set up, and try a few things you didn't — then come back to the chat with anything you'd like handled differently.
  </Step>

  <Step title="Fine-tune on the other tabs">
    When the shape is right, polish the rest by hand: the agent's [personality and voice](/building/personality-and-voice), the [journeys](/building/journeys) in detail, and the [knowledge it answers from](/building/agent-knowledge).
  </Step>
</Steps>

<Note>
  Everything you build here is saved to your agent's draft as you go — nothing changes for real callers until you [publish](/building/publishing).
</Note>