llm-router

What This Skill Does

The llm-router skill acts as a unified gateway for OpenClaw, allowing your autonomous agents to access over 70 different AI models through a single, OpenAI-compatible API interface. By integrating this skill, you eliminate the overhead of managing dozens of individual API keys, disparate SDKs, and fragmented billing cycles. Powered by the AIsa infrastructure, the router supports major model families including GPT-4o, Claude 3.5 Sonnet, Gemini 2.5, and Grok. It simplifies agent architecture by enabling dynamic model switching and sophisticated fallback strategies, ensuring that your agents can leverage the most appropriate intelligence for any given task.

Installation

To integrate the llm-router into your OpenClaw environment, execute the following command in your terminal:

clawhub install openclaw/skills/skills/chaimengphp/openclaw-aisa-llm-gateway-route

After installation, ensure you have set your global authentication key by exporting your credential: export AISA_API_KEY="your-key". The skill will automatically configure the endpoint to point toward the unified AIsa gateway.

Use Cases

This skill is designed for high-performance automation workflows. Use it to implement multi-model reasoning where an agent performs an initial analysis with a fast model like 'gpt-4o-mini' and routes complex, high-reasoning tasks to 'claude-opus-4'. It is also essential for developers conducting model comparison testing or building resilient agents that utilize built-in fallback strategies; if your primary choice is unavailable due to rate limits or outages, the router can seamlessly failover to a secondary provider without interrupting the agent's logic.

Example Prompts

1. "Analyze the current market trends from the provided document using GPT-4, then summarize the insights with Claude-3.5."
2. "Compare the creative writing output of Grok-3 and Gemini-2.5-pro for this specific science-fiction prompt."
3. "If GPT-4o fails to return a valid JSON object for this dataset, automatically retry the query using Llama-3.1-405b."

Tips & Limitations

To optimize costs, prioritize 'mini' or 'flash' variants for simple chat tasks and reserve premium 'opus' or 'pro' models for deep reasoning. Keep in mind that model availability is dynamic; always check the official AIsa pricing and availability documentation if you notice specific models returning errors. Ensure your agent environment variables are correctly scoped to prevent leakage of the single unified API key, as it provides access to all your configured providers simultaneously.