claudish-usage

Claudish CLI enables running Claude Code with external models like Grok, GPT-5, Gemini, and MiniMax via OpenRouter. Always delegate Claudish tasks to a sub-agent rather than executing directly in your main context to prevent context window pollution and maintain clean agent isolation.

Route all Claudish and external-model tasks through sub-agents by default. Run Claudish directly in main context only when you explicitly request it. Sub-agent delegation prevents context contamination, maintains workflow clarity, and follows the mandatory delegation pattern for OpenRouter proxy usage.

Use claudemem status to check available models, then claudemem search to find model recommendations matching your task needs. The agent selection matrix in Claudish guides routing for coding, review, and architecture work across OpenRouter's available models.

Claudish supports file-based instructions that keep prompts lean and maintain context efficiency. Instead of embedding large directives inline, reference external instruction files to reduce token consumption while preserving full workflow context in sub-agents.

Yes. Claudish CLI routes to OpenRouter proxy, enabling access to Grok, GPT-5, Gemini, MiniMax, and other external models. Always delegate these cross-model tasks to sub-agents and follow agent selection guidance for optimal routing based on task type.

Running Claudish directly in main context risks context window pollution and violates the mandatory sub-agent delegation pattern. Claudish is designed for sub-agent execution only; direct execution degrades isolation, reduces context efficiency, and breaks the intended workflow decision tree.