Free Claude Code

Free Claude Code is an open-source local proxy server that lets you run Claude Code, Codex CLI, and Pi using any model provider you choose. With over 43,000 GitHub stars and 7,000 forks, it has become one of the most popular ways to access AI coding agents without being locked into a single provider's subscription.

Rather than replacing Claude Code or Codex, FCC sits as a transparent proxy between the agent and its model provider. You keep the agent's native interface and feature set — streaming, tool use, reasoning, image input — while routing traffic through providers you already have access to. The project is MIT-licensed and hosted on GitHub at github.com/alishahryar1/free-claude-code.

FCC is written in Python and distributed via the uv package manager. It runs as a background server with a local Admin UI for configuration, available on Windows, macOS, and Linux.

• Launch Any Agent - Run Claude Code (fcc-claude), Codex (fcc-codex), or Pi (fcc-pi) with your own provider.
• 31+ Providers - Switch among cloud providers (OpenAI, Google, DeepSeek, OpenRouter, NVIDIA NIM) and local models (Ollama, LM Studio, llama.cpp).
• Native Model Picker - Each agent's built-in model selector works with FCC-exposed models.
• Full Feature Passthrough - Streaming, tool use, reasoning, and image input all work with compatible providers.
• Desktop Integration - Runs in system tray (Windows) or menu bar (macOS) with a launcher shortcut.
• Admin UI - Local browser interface for managing providers, models, and proxy settings.
• Chat Integrations - Optionally run Claude Code sessions through Discord or Telegram with voice-note transcription.
• Token Auth - Optional token-based authentication to protect the local proxy.

macOS / Linux:

curl -fsSL "https://raw.githubusercontent.com/Alishahryar1/free-claude-code/main/scripts/install.sh" | sh

Windows PowerShell:

& ([scriptblock]::Create((irm "https://raw.githubusercontent.com/Alishahryar1/free-claude-code/main/scripts/install.ps1")))

Re-run the same command to update. On Windows and macOS, FCC installs as a desktop launcher. On Linux, run fcc-server from the terminal. To check the installed version: fcc-server --version.

Starting FCC

On Windows and macOS, open "Free Claude Code" from the Start menu or Applications folder — it runs in the system tray or menu bar. On Linux, run fcc-server in the terminal. The Admin UI opens automatically in your browser at http://127.0.0.1:8082/admin once the server is healthy.

FCC supports over 31 cloud and local providers. You configure each in the Admin UI with an API key or base URL, then select a model from the dropdown. Key providers include:

Provider	Admin Setting	Example Model
NVIDIA NIM	NVIDIA_NIM_API_KEY	nvidia_nim/nvidia/nemotron-3-super-120b
OpenAI / ChatGPT	Connect ChatGPT in Admin	openai/<model-id>
OpenRouter	OPENROUTER_API_KEY	open_router/openrouter/free
Google AI Studio	GEMINI_API_KEY	gemini/models/gemini-3.1-flash-lite
DeepSeek	DEEPSEEK_API_KEY	deepseek/deepseek-chat
GitHub Models	GITHUB_MODELS_TOKEN	github_models/openai/gpt-4.1
Kilo.ai	KILO_API_KEY	kilo/kilo-auto/free
Groq	GROQ_API_KEY	groq/llama-3.3-70b-versatile
Ollama (local)	OLLAMA_BASE_URL	ollama/<model-tag>
LM Studio (local)	LM_STUDIO_BASE_URL	lmstudio/<model-id>
OpenCode Zen	OPENCODE_API_KEY	opencode/gpt-5.3-codex
Hugging Face	HUGGINGFACE_API_KEY	huggingface/Qwen/Qwen3-Coder-480B

OpenAI uses your ChatGPT subscription via browser PKCE authentication — no API key needed. Local providers (Ollama, LM Studio, llama.cpp) let you run models entirely offline without any cloud dependency.

Once FCC is running and you have configured a provider, launch your coding agent with one of three commands:

Agent	Command	Description
Claude Code	fcc-claude	Anthropic's agent with native model picker showing FCC providers
Codex CLI	fcc-codex	OpenAI's agent routed through your chosen provider
Pi	fcc-pi	Minimalist coding agent via FCC proxy

All three launchers use the current Admin UI settings. Normal CLI arguments still work — for example, fcc-codex exec "hello". The fcc-pi launcher registers FCC only for that Pi process, leaving your existing Pi settings unchanged.

The Admin UI is a local browser interface where you manage all FCC settings. It opens automatically at http://127.0.0.1:8082/admin when the server starts. From the Admin UI you can:

• Configure Providers — Enter API keys, base URLs, and select models for each provider.
• Route Traffic — Map Claude Code's native model names (Fable, Opus, Sonnet, Haiku) to any FCC provider model.
• Set Fallbacks — Configure fallback providers so your agents continue working if the primary provider is unavailable.
• Token Auth — Enable optional token authentication to restrict who can use the proxy on your network.
• Monitor Status — See server health, connected agents, and active provider connections at a glance.

On Windows, left-clicking the system tray icon opens Admin directly. On macOS, use the menu bar icon. Changes take effect immediately — no server restart required for most settings.

Desktop Environment

FCC integrates into your desktop as a background server. On Windows and macOS, it runs in the system tray or menu bar with a dedicated launcher installed during setup. The server starts automatically and you can access the Admin UI, check server status, restart, or quit from the tray/menu. On Linux, it runs in the terminal.

VS Code / JetBrains

You can connect Claude Code and Codex within VS Code through FCC, or use Claude Code through JetBrains ACP (Agent Client Protocol). This gives you the native IDE experience while routing model traffic through your configured providers.

Chat Platforms

Optionally run Claude Code sessions through Discord or Telegram with voice-note transcription. This allows you to interact with your coding agents from your phone or messaging platform.

Local Models

For fully offline usage, FCC supports local providers such as Ollama, LM Studio, and llama.cpp. You can run coding agents entirely on your own hardware with no internet connection or API costs.