crossmodel
claudeDelegate work from Claude Code to any external LLM CLI (Codex, Grok, Cursor, OpenCode — and through it OpenRouter, Ollama and any OpenAI-compatible endpoint — or your own), spending that provider's quota instead of your Anthropic quota. Includes a deterministic multi-provider benchmark to decide what each model can be trusted with, and a permission policy that lets an agent write and test while git history stays yours.
2 skills · 3 agents · hooks
1