Works with LiteLLM

LiteLLM, running on your Mac

ToolPiper is a LiteLLM provider, so every tool that routes through LiteLLM reaches a local model without knowing anything about ToolPiper.

Download ToolPiper Read the API docs

Free, Apple Silicon, signed. The local model runner has no caps and needs no account.

Setup

  1. Install ToolPiper and launch it once. The local API listens on port 9998.
  2. Prefix the model with `toolpiper/`.
from litellm import completion

completion(
    model="toolpiper/qwen3.8-27b",
    messages=[{"role": "user", "content": "Hello"}],
)

Prompts, responses and files stay on the machine. There is no key to rotate and no per-token cost. LiteLLM talks to ToolPiper over its local OpenAI-compatible API on port 9998.