Recipes
Short, complete setups for pointing an AI coding assistant at a model you control. Each one is a full walkthrough: what you need, the steps, and what it costs.
Run Claude Code on Apple Intelligence
Free, on-devicePoint Claude Code at Apple Intelligence on your Mac's Neural Engine. No API key, no cloud round-trip, no per-token billing.
Run Claude Code on Local Qwen
Free, on-deviceWire Claude Code to a local Qwen model through llama.cpp. The best balance of capability and privacy: long context, tool use, and your code and prompts stay on the Mac.
Use Claude Code with your OpenAI key
Your own API billingBring your own OpenAI key. ToolPiper injects it server-side from the macOS Keychain, so it never lands in a dotfile or a log.
Switch Claude Code providers in chat
MixedConfigure providers once, then change backend from inside Claude Code in natural language. No env vars, no restart.