Documentation
Everything you need to install, use, and create AI agent souls and skills.
⚡ Quick Start
1. Install a free skill
/install crazyrouter-chat2. Try it
Use Claude to explain quantum computing in one paragraph3. Want more? Install a soul
/install xiaohongshu-agent🚀Getting Started
What are Souls & Skills?▼
A soul is a complete AI agent identity — persona, memory, decision principles, and skills bundled together. A skill is a single capability (like image generation or translation) that any agent can use.
Think of it this way: a soul is who the agent is. A skill is what the agent can do.
Prerequisites▼
You need:
• An AI bot running on OpenClaw, or any compatible gateway
• A Crazyrouter API key (free tier available at crazyrouter.com)
• That's it. Seriously.
⚡Installation
Bot Command▼
The fastest way. In your bot chat:
/install <name>
Examples:
• /install crazyrouter-chat — Free, adds 627+ model access
• /install xiaohongshu-agent — $29, Xiaohongshu content expert
• /install content-team — $99, 3 coordinated agents
CLI▼
For developers who prefer the terminal:
npx crazyrouter-souls install <name>
npx crazyrouter-souls list
npx crazyrouter-souls info <name>
npx crazyrouter-souls uninstall
What Gets Installed▼
Each package includes:
• SOUL.md — Agent identity and personality
• MEMORY.md — Pre-loaded knowledge and context
• skills/*.md — Capabilities with API configs
• All pre-configured with Crazyrouter API
🛠️Creating & Selling
Create Your Own Soul▼
Structure your soul package:
my-soul/
├── SOUL.md # Identity
├── MEMORY.md # Knowledge
├── skills/
│ ├── skill-1.md # Capability
│ └── skill-2.md # Capability
└── README.md # Documentation
Test it locally, then submit to our marketplace.
Submission Guidelines▼
• Must be tested in production (not just theory)
• Include clear documentation
• Specify which models/APIs it uses
• Set a fair price (or make it free)
• We review within 48 hours
📡API Reference
Crazyrouter API▼
All souls and skills use Crazyrouter API as the default provider.
Base URL: https://crazyrouter.com/v1
OpenAI-compatible. Drop-in replacement for any OpenAI SDK.
627+ models including GPT-4o, Claude, Gemini, DeepSeek, Llama, Mistral, and more.
Endpoints▼
• POST /v1/chat/completions — Text generation
• POST /v1/images/generations — Image generation
• POST /v1/audio/speech — Text to speech
• POST /v1/audio/transcriptions — Speech to text
• GET /v1/models — List available models
Ready to get started?