---
title: PraisonChat
emoji: 🐱
colorFrom: purple
colorTo: blue
sdk: docker
app_port: 7860
pinned: false
---

# PraisonChat 🐱

An **OpenClaw-style** multi-agent AI chat powered by **PraisonAI** and **LongCat Flash Lite**.

## Features
- 🤖 Dynamic sub-agent creation — the main agent spawns specialized agents per task
- 🔧 Auto tool creation — agents write their own Python tools on the fly
- 💬 Streaming responses with real-time agent step visibility
- 📝 Full markdown + code highlighting
- 🌓 Dark/light mode
- 💾 Local chat history
- 🎯 Multi-model support (Flash Lite, Flash Chat, Flash Thinking)

## Setup
1. Get a free LongCat API key at [longcat.chat/platform](https://longcat.chat/platform)
2. Enter your API key in Settings (⚙️)
3. Start chatting!

## Models
| Model | Context | Speed | Free Quota |
|-------|---------|-------|------------|
| LongCat Flash Lite | 320K | Fastest | 50M tokens/day |
| LongCat Flash Chat | 256K | Fast | 500K tokens/day |
| LongCat Flash Thinking | 256K | Medium | 500K tokens/day |

## Environment Variables (optional)
Set `LONGCAT_API_KEY` as a Space Secret to pre-configure the API key for all users.

## Architecture
```
User Message
    │
    ▼
Main Orchestrator Agent (analyzes task, plans execution)
    │
    ├── Creates Sub-Agent 1 with custom tools
    ├── Creates Sub-Agent 2 with custom tools
    └── Creates Sub-Agent N with custom tools
         │
         ▼
    Each agent executes its task
         │
         ▼
    Orchestrator synthesizes → Final response to user
```