Introduction

Fastlane is a local proxy that learns which LLM works best for each type of prompt and routes accordingly. It sits between your tools and the APIs, saving cost while improving quality.

What Fastlane Does

• Routes intelligently - A trained ranking model picks the best provider for each request
• Saves money - Simple prompts go to cheap models, complex ones to strong models
• Keeps data private - Everything runs on your machine, no data leaves unless you send it to a provider
• Works everywhere - Drop-in replacement for any OpenAI-compatible endpoint

How It Works

Your App → Fastlane Proxy → Best Provider
     ↑                           ↓
     ←←←←← Response ←←←←←←←←←←←

Fastlane runs a local server at 127.0.0.1:8787 that speaks the OpenAI API format. Point your tools at it and they automatically get smart routing.

Key Features

• ONNX ranking model - Trained on prompt complexity, runs locally
• 5 provider backends - OpenAI, Anthropic, Groq, OpenRouter, Marketplace
• 3 routing modes - Cheap, Balanced, Accurate
• No login required - Just install and go
• Bring Your Own Keys - Use your existing API keys
• Marketplace option - Top up via Razorpay, we handle the rest