---
title: For LLMs
description: Humation is small enough for a coding agent to use correctly from a short brief. This page is that brief.
---

## llms.txt

A machine-readable summary of the packages, the quick start, and the rules below lives at [/llms.txt](/llms.txt). Point an agent at that URL rather than pasting code samples.

## Rules worth stating explicitly

These are the points agents most often get wrong when they assume Humation works like a hosted avatar service.

```text
Use @humation/react for React UI.
Use @humation/assets-humation-1 for the asset set. It is always required.
Do not call a Humation API or any remote URL for rendering.
Use seed for deterministic generated avatars.
Use selections and colors for user-customized avatars.
Use @humation/core only when raw SVG output or picker previews are needed.
Part names are scoped to a slot: { head: 'braids' }, not { part: 'braids' }.
Empty slots are the part named 'none', not undefined.
```

## Minimal correct example

If an agent produces only this much, it is already correct. Everything else is refinement.

```tsx
import { Avatar } from '@humation/react';
import { humation1 } from '@humation/assets-humation-1';

<Avatar assets={humation1} seed={user.id} size={96} title={user.name} />;
```

The full surface is on [Options](/options), and every part name is on [Parts](/parts). Both pages are short enough to paste into a prompt whole. For stored data, remind the agent to use canonical IDs such as `hm1-p-000020`.