---
title: "S&box AI Helper - s&box library"
description: "An AI chatbot integrated directly into the S&box editor. Chat with a local AI assistant trained on S&box"
canonical: "https://sbox.watch/libraries/icu/ai-chat/"
generated_at: "2026-07-22T23:21:23.098Z"
---

# S&box AI Helper

An AI chatbot integrated directly into the S&box editor. Chat with a local AI assistant trained on S&box

- Type: library
- Ident: icu.ai-chat
- Creator: icu
- Canonical: https://sbox.watch/libraries/icu/ai-chat/
- Markdown: https://sbox.watch/libraries/icu/ai-chat.md
- sbox.game: https://sbox.game/icu/ai-chat/
- Generated: 2026-07-22T23:21:23.098Z

## Stats

- 24h momentum: +0
- Terry score: 35%
- Favorites: 3
- Upvotes: 4
- Downvotes: 2
- Updated: Apr 17, 2026
- Created: Apr 17, 2026
- Size: 41.4 MB
- Files: 268
- Tags: ai, ai, bot, chat, helper, tools

## Description

\# AI Chat for S&box

An AI chatbot integrated directly into the S&box editor. Chat with a local AI assistant while you build your game.

\#\# Features
- Dockable editor panel — dock it anywhere in your layout
- Uses \*\*Ollama\*\* for free, local AI (no API key needed)
- Supports any Ollama model (llama3, mistral, phi3, etc)
- Conversation history maintained during session
- Customisable system prompt

\#\# Setup

\#\#\# 1. Install Ollama
Download from \[ollama.com\](https://ollama.com) and install it.

\#\#\# 2. Pull a model
\`\`\`
ollama pull llama3
\`\`\`

\#\#\# 3. Start Ollama on port 8080
S&box only allows localhost on specific ports. Start Ollama with:

\*\*CMD:\*\*
\`\`\`
set OLLAMA\_HOST=0.0.0.0:8080 && ollama serve
\`\`\`

\*\*PowerShell:\*\*
\`\`\`
$env:OLLAMA\_HOST="0.0.0.0:8080"; ollama serve
\`\`\`

\#\#\# 4. Open AI Chat in S&box
Find it in \*\*View > AI Chat\*\* or look for the AI Chat panel in your dock options.

\#\# Usage
- Type a message in the input field
- Press \*\*Enter\*\* or click \*\*Send\*\*
- The AI responds using your local Ollama model
- Change the model name in the field at the top

\#\# Notes
- Ollama must be running on port 8080 before using the chat
- All processing is local — nothing is sent to external servers
- Larger models give better responses but are slower

\#\# Credits
Made by ICU