Free Claude Code is a Python-based proxy system that enables users to route Claude Code and Codex API calls through alternative AI model providers, effectively allowing free or self-hosted access to these Anthropic and OpenAI tools. The repository has accumulated 38,571 stars as of its most recent tracking update, with steady growth of 3 stars since the previous check on July 4, 2026. The project is maintained primarily by Alishahryar1, who has logged 308 tracked events, with secondary contributors AAsdfd3253 and LVT382009 contributing 22 and 21 events respectively. The repository shows active community engagement with a median issue and pull request response latency of 1.6 hours across 332 tracked items, though the mean response time extends to 28.5 hours, indicating variable response patterns depending on issue complexity.

The core functionality centers on intercepting and redirecting API traffic from Claude Code's CLI and VS Code extension, which normally communicate with Anthropic's Messages API, as well as Codex traffic using OpenAI's Responses API. Rather than forcing users to choose a single provider, Free Claude Code maintains protocol stability for each client while allowing flexible routing to 18 different backend providers including NVIDIA NIM, OpenRouter, Google AI Studio Gemini, DeepSeek, Mistral La Plateforme, Mistral Codestral, Cerebras Inference, Groq, Fireworks AI, Cloudflare, LM Studio, llama.cpp, and Ollama. This architecture enables users to leverage free tier offerings, paid alternatives, or entirely local models through a unified proxy interface.

The system provides launcher commands fcc-claude and fcc-codex that dynamically read the current Admin UI port and authentication token on each startup, eliminating the need for manual configuration updates. Users can implement per-model routing for Claude Code, directing Opus, Sonnet, Haiku, and fallback traffic to different providers based on their needs. The proxy supports native model picker integration, allowing Claude Code users to select from available models through the standard /model command, while Codex users access a generated local model catalog. Advanced features include streaming support, tool use handling, reasoning and thinking block processing, and optional voice-note transcription via local Whisper or NVIDIA NIM.

The repository includes optional integrations for Discord and Telegram bot wrappers enabling remote Claude Code sessions, support for the Claude Code VS Code extension, and Codex CLI and VS Code extension compatibility through shared configuration files. A local Admin UI accessible at the /admin endpoint provides loopback-only access for editing proxy settings, validating configuration changes, and checking provider status without exposing controls to remote access.

Installation is streamlined through platform-specific scripts for macOS, Linux, and Windows PowerShell that handle dependency management including the uv package manager and Python runtime. The quick start process involves installing the proxy, starting the server, configuring a provider through the Admin UI, and launching either fcc-claude or fcc-codex to begin using the routed services. The repository's classification spans Claude API integration, AI chatbot functionality, code generation, natural language processing, prompt engineering, automation, and machine learning applications. Related projects tracked through overlapping contributors include ruvnet/ruflo, nousresearch/hermes-agent, and openclaw/openclaw, suggesting an ecosystem of complementary AI tooling and integration projects.