A comprehensive CLI integration for Openclaw Skills that enables seamless interaction with 627+ AI models through the Crazyrouter API.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install crazyrouter-chat
2. Manual Installation
Copy the skill folder to one of these locations
~/.openclaw/skills/ <project>/skills/ Priority: Workspace > Local > Bundled
3. Prompt Installation
Copy this prompt to OpenClaw to install it automatically.
Help me install crazyrouter-chat using Clawhub. If Clawhub is not installed, install it first (npm i -g clawhub).
Prefer to download?
Get the raw skill files in a ZIP archive.
What is crazyrouter-chat?
Crazyrouter Chat is a powerful tool designed to consolidate access to the vast landscape of artificial intelligence. By utilizing the Crazyrouter API, this skill allows developers to communicate with a massive array of models—including frontier ones like GPT-5, Claude Opus 4.6, and DeepSeek R1—using a single, unified interface. It is an essential addition to your library of Openclaw Skills, providing the flexibility to switch providers without changing your underlying code or managing multiple subscriptions.
Whether you need to benchmark different model performances, generate specialized code, or simply seek a second opinion from a specific AI architecture, this skill simplifies the process. It handles the routing and communication logic, allowing you to focus on the prompt engineering and data application. This is a technical, developer-centric solution for those who require high-velocity access to the latest AI advancements through a clean CLI workflow.
crazyrouter-chat Use Cases
- Running side-by-side comparisons of model responses to evaluate accuracy and logic.
- Querying specialized models like DeepSeek R1 for complex programming tasks or Grok 4 for real-time information.
- Automating the generation of README files or documentation directly to the local file system.
- Testing system prompts across multiple LLM providers to ensure persona consistency.
- Debugging code by getting diverse perspectives from different AI model families.
How crazyrouter-chat Works
- The user executes a command via the CLI, specifying the desired model and prompt.
- The skill performs a blocking check to ensure the CRAZYROUTER_API_KEY environment variable is configured.
- The request is processed by the internal TypeScript script using the Bun runtime for high performance.
- The prompt and parameters (like temperature and max tokens) are sent to the Crazyrouter gateway.
- Crazyrouter routes the request to the specific provider (OpenAI, Anthropic, Google, etc.).
- The response is received and either printed to the terminal or saved to a designated output file.
crazyrouter-chat Setup
To integrate this into your Openclaw Skills, follow these steps:
- Obtain your API key from crazyrouter.com.
- Set the environment variable in your shell:
export CRAZYROUTER_API_KEY="your_actual_key_here"
- Verify the key is active:
echo "${CRAZYROUTER_API_KEY:-not_set}"
- Use the skill via npx and bun:
npx -y bun scripts/main.ts --model gpt-5 --prompt "Hello World"
crazyrouter-chat Data Schema & Taxonomy
The skill manages data primarily through command-line arguments and file outputs. It supports structured interactions using the following schema:
| Parameter | Role | Default |
|---|---|---|
--prompt |
The primary text input for the AI | Required |
--model |
The specific model ID (e.g., gpt-5, claude-sonnet-4) | gpt-5-mini |
--system |
Defines the behavior or persona of the AI | Optional |
--output |
The file path where the response will be persisted | Console only |
--json |
Returns the raw API response for programmatic parsing | False |
All metadata regarding the request, including token usage and model timestamps, is available when using the --json flag.
crazyrouter-chat Advanced Features
- Multi-model benchmarking: Easily swap model IDs to compare output quality across 627+ options.
- Persona persistence: Use the system prompt flag to maintain specific constraints across different session triggers.
- Output redirection: Directly write AI-generated code or text to files for automated pipeline integration.
- Fine-tuned sampling: Control creativity and deterministic behavior using the temperature argument.
- Dynamic model discovery: Use the list-models flag to stay updated on the newest available architectures in the Openclaw Skills ecosystem.
SKILL.md
Loading
Related Openclaw Skills
Chinese Holiday Checker Skill
A high-precision tool for AI agents to identify Chinese statutory holidays and compensatory workdays based on the latest official schedules.
A-Stock Quantitative Monitor & Selector
A comprehensive A-share quantitative system providing real-time monitoring, 7-dimension market sentiment scoring, and multi-strategy stock selection.
OpenClaw Router
A specialized skill for installing, updating, and maintaining the core OpenClaw Router plugin for seamless gateway management.
2fun-pansou: Chinese Cloud Drive Search
A search aggregator for discovering media and files across major Chinese cloud storage platforms.
ArXiv Batch Reporter
An automation tool that aggregates individual ArXiv paper summaries into structured, professional collection reports using template-based injection.
ArXiv Search Collector
A model-driven workflow for intelligent arXiv paper retrieval, multi-query expansion, and automated relevance filtering.
Featured*
