Do I need separate API keys for each model?

No, you only need one OpenRouter API key set as an environment variable to access all supported models through the Claudish interface.

Why does this skill mandate using sub-agents?

Running external models directly in the main context can add over 10,000 tokens of reasoning and metadata, quickly exhausting your context window. Sub-agents isolate this data.

What is the file-based instruction pattern?

It is a best practice where instructions are written to a temporary file and passed to Claudish, ensuring the sub-agent has clear requirements without cluttering the main terminal session.

Which AI models are recommended for coding tasks?

The skill recommends x-ai/grok-code-fast-1 for speed, openai/gpt-5 for complex reasoning, and google/gemini-2.5-flash for massive context tasks.

What is Claudish and how does it work with Claude Code?

Claudish is a proxy tool that allows Claude Code to interface with any OpenRouter-compatible model by routing requests through a local server that mimics the Anthropic API.

Claudish OpenRouter Integration

Name: Claudish OpenRouter Integration
Author: involvex

byinvolvex

Developer Tools

Integrates OpenRouter models like Grok, GPT-5, and Gemini into Claude Code using specialized sub-agent delegation to preserve context.

About

Claudish Usage provides a standardized framework for running Claude Code with external models via OpenRouter while maintaining context window efficiency. By enforcing a sub-agent-only execution pattern, it prevents main context pollution and allows developers to leverage specialized models like Grok for speed, GPT-5 for reasoning, or Gemini for long-context tasks without degrading the performance of the primary AI session. This skill includes a comprehensive agent selection matrix and file-based instruction patterns to ensure production-grade task execution across multiple AI providers.

Key Features

File-based instruction patterns for complex, multi-step AI workflows
0 GitHub stars
Intelligent agent selection matrix for specialized coding and architecture tasks
Support for direct agent invocation via the --agent CLI flag
Seamless access to OpenRouter models including Grok, GPT-5, and Gemini
Mandatory sub-agent delegation to prevent main context window pollution

Use Cases

Delegating heavy code reviews to specialized reasoning models like GPT-5
Executing large-scale codebase analysis with Gemini’s million-token context window
Performing fast, iterative code implementations using Grok-code-fast

About

Key Features

File-based instruction patterns for complex, multi-step AI workflows
0 GitHub stars
Intelligent agent selection matrix for specialized coding and architecture tasks
Support for direct agent invocation via the --agent CLI flag
Seamless access to OpenRouter models including Grok, GPT-5, and Gemini
Mandatory sub-agent delegation to prevent main context window pollution

Use Cases

Delegating heavy code reviews to specialized reasoning models like GPT-5
Executing large-scale codebase analysis with Gemini’s million-token context window
Performing fast, iterative code implementations using Grok-code-fast