Crazyrouter Chat

Chat with 627+ AI models via Crazyrouter. GPT-5, Claude Opus 4.6, Gemini 3, DeepSeek R1, Llama 4, Qwen3, Grok 4. Use when user wants to query a specific model, compare model responses, or get a second opinion from a different AI. Requires environment variable CRAZYROUTER_API_KEY (get at https://crazyrouter.com).

Audits

Malicious

ClawScanPending

Agentic behavior and permission review.

Static analysisMalicious

Pattern checks against bundled files.

VirusTotalPending

Multi-engine malware detections and file reputation.

Install

openclaw skills install crazyrouter-chat

Multi-Model Chat via Crazyrouter

Chat with any of 627+ AI models through Crazyrouter — one API key, all providers.

Supported Models (highlights)

Provider	Models
OpenAI	gpt-5, gpt-5-mini, gpt-4.1, gpt-4o, o3, o4-mini
Anthropic	claude-opus-4-6, claude-sonnet-4, claude-haiku-3.5
Google	gemini-3-pro, gemini-2.5-flash
DeepSeek	deepseek-r1, deepseek-v3
Meta	llama-4-scout, llama-4-maverick
Alibaba	qwen3-235b, qwen3-32b
xAI	grok-4, grok-3
Mistral	mistral-large, codestral

Script Directory

Agent Execution:

SKILL_DIR = this SKILL.md file's directory
Script path = ${SKILL_DIR}/scripts/main.mjs

Step 0: Check API Key ⛔ BLOCKING

[ -n "${CRAZYROUTER_API_KEY}" ] && echo "key_present" || echo "not_set"

Result	Action
`key_present`	Continue
`not_set`	Ask user to set `CRAZYROUTER_API_KEY`. Get key at https://crazyrouter.com

Usage

# Ask GPT-5 a question
node ${SKILL_DIR}/scripts/main.mjs --model gpt-5 --prompt "Explain quantum computing in 3 sentences"

# Compare models
node ${SKILL_DIR}/scripts/main.mjs --model deepseek-r1 --prompt "Write a sorting algorithm in Python"
node ${SKILL_DIR}/scripts/main.mjs --model claude-opus-4-6 --prompt "Write a sorting algorithm in Python"

# With system prompt
node ${SKILL_DIR}/scripts/main.mjs --model gemini-3-pro --system "You are a pirate" --prompt "Tell me about the weather"

# With temperature
node ${SKILL_DIR}/scripts/main.mjs --model gpt-5-mini --prompt "Write a poem" --temperature 1.2

# Save output to file
node ${SKILL_DIR}/scripts/main.mjs --model gpt-5 --prompt "Write a README" --output readme.md

# List available models
node ${SKILL_DIR}/scripts/main.mjs --list-models

Options

Option	Description	Default
`--prompt <text>`	User message (required)	—
`--model <id>`	Model to use	`gpt-5-mini`
`--system <text>`	System prompt	—
`--temperature <num>`	Sampling temperature (0-2)	—
`--max-tokens <num>`	Max response tokens	—
`--output <path>`	Save response to file	—
`--list-models`	List popular models	—
`--json`	Output raw JSON	—