Free Claude Code is a proxy server that lets you use Claude Code without an Anthropic subscription. It intercepts the API calls Claude Code makes and routes them through your choice of six provider backends: NVIDIA NIM (free tier available), OpenRouter, DeepSeek, LM Studio, llama.cpp, and Ollama.
Setting it up takes under five minutes. Clone the repo, configure your .env file with your chosen provider, start the proxy with uvicorn, and point Claude Code at localhost:8082. The proxy handles streaming, tool use, reasoning blocks, and all the Anthropic protocol features Claude Code expects.
Where Free Claude Code shines is per-model routing. You can send Opus-tier requests to OpenRouter, Sonnet requests to DeepSeek, and Haiku requests to a local Ollama instance. This lets you balance cost and quality across different coding tasks. The fallback MODEL setting catches any unspecified tiers.
Beyond the terminal, Free Claude Code works with the VS Code extension and JetBrains ACP. Configure environment variables in your IDE settings and the proxy handles the rest. A model picker script (claude-pick) lets you choose your model at launch time.
Optional Discord and Telegram bot wrappers let you run Claude Code sessions remotely. Stream coding progress to a channel, reply to branch conversations, and stop or clear tasks from chat. Voice note transcription through Whisper or NVIDIA NIM adds hands-free interaction.
The project has 18.7k GitHub stars, 547 commits, and active development with security fixes and provider updates. It supports Python 3.14 and uses uv for fast dependency management. The MIT-licensed codebase is well-documented with per-provider setup guides and troubleshooting sections.