Overview
Unlock the full power of Claude Code without the cost of API keys. This project provides a lightweight, transparent proxy that seamlessly routes your Claude Code CLI and VSCode extension requests to a variety of powerful LLM providers, including NVIDIA NIM (generous free tier), OpenRouter, DeepSeek, and even fully local solutions like LM Studio and llama.cpp. It's a drop-in replacement, requiring minimal configuration to start enjoying free, advanced AI coding assistance.
Key Features
- Zero Cost Access: Utilize NVIDIA NIM's 40 free requests per minute, explore a vast array of models on OpenRouter, or run powerful AI models entirely locally with LM Studio or llama.cpp.
- Seamless Integration: Acts as a direct replacement for Anthropic's API. Simply set two environment variables (
ANTHROPIC_BASE_URLandANTHROPIC_AUTH_TOKEN), and your existing Claude Code setup will work out-of-the-box. - Flexible Provider Routing: Map Claude's Opus, Sonnet, and Haiku models to different providers and specific LLM models. Mix and match providers to optimize for cost, performance, or model capabilities.
- Enhanced AI Capabilities: Automatically parses
<think>tags andreasoning_contentinto native Claude thinking blocks. Features a heuristic tool parser to convert text-based tool calls into structured formats. - Intelligent Request Optimization: Intercepts and handles common, trivial API calls locally, saving your valuable API quota and reducing latency.
- Robust Rate Limiting: Implements smart, multi-layered rate limiting including rolling window throttling, exponential backoff for 429 errors, and optional concurrency caps.
- Discord/Telegram Bot Integration: Control Claude Code remotely from your favorite chat platforms, with advanced features like tree-based threading, session persistence, and live progress streaming.
- Voice Note Support: Send voice messages to your bot, which will be transcribed and processed as regular prompts using local Whisper or NVIDIA NIM backends.
Typical Use Cases
- Developers: Integrate free, high-performance AI coding assistance directly into your terminal and VSCode workflow without incurring API costs.
- Hobbyists & Students: Experiment with advanced AI models for coding, text generation, and more, with no financial barrier.
- AI Enthusiasts: Explore the capabilities of various LLM providers, including local inference, by routing them through a familiar Claude Code interface.
- Remote Collaboration: Manage and deploy AI coding tasks remotely via Discord or Telegram bots, ideal for team projects or personal assistants.
- Offline Development: Leverage powerful local LLMs for privacy-conscious coding or in environments without internet access.
