Free Claude Code

Overview

Free Claude Code (FCC) is an independent open-source project that routes free, paid, subscription, and local models through a local proxy and Admin UI. It supports 61 providers and 12 coding agents, letting users keep their preferred agent across terminal, desktop, IDE, browser, and messaging clients.

Key Features

• Supports 61 ToS-friendly providers and more than 1.3 billion free tokens per month in one searchable UI.
• Runs 12 coding agents, including Claude Code, Codex, Pi, OpenCode, Cline, Aider, and VS Code Chat, against a shared model catalog.
• Automatically tries the next configured model after retries are exhausted, across every client.
• Offers native Codex browser sessions with real-time and background support and provider or model switching within a session.
• Provides optional RTK output filtering and Headroom tool-result compression to reduce token usage by up to 90%.
• Supports voice notes through local Whisper or NVIDIA NIM transcription and preserves streaming, tools, images, and interleaved thinking.

Use Cases

• Route coding agents through free-tier, subscription, or local models from one shared catalog.
• Keep a coding session running when a provider fails by configuring fallback models.
• Access coding agents from a terminal, IDE, desktop app, browser session, Discord, or Telegram.
• Use voice notes to prompt a coding agent with local Whisper or NVIDIA NIM transcription.

Getting Started

• Install using the provided macOS/Linux shell script or Windows PowerShell installer.
• Start FCC from the desktop or Start menu on Windows, from Applications on macOS, or by running fcc-server on Linux and keeping the terminal open.
• Open the Admin UI, add a provider key such as NVIDIA NIM, select a model, and click Apply.
• Run a launcher such as fcc-claude, fcc-codex, fcc-pi, or fcc-opencode.

Deployment & Requirements

• Installers are provided for macOS/Linux and Windows PowerShell.
• Provider credentials, such as an API key or OAuth connection, are configured in the Admin UI before selecting models; NVIDIA NIM is the default example.
• Local model use requires a running LM Studio, llama.cpp, or Ollama server and tool-capable models with sufficient context.
• Editor and app integrations require FCC to run on the same computer and in the same user environment as the client.

Before You Adopt

• License: Other. Review its terms before using, modifying, or distributing the project.
• Free-tier availability and limits are set by each provider and can change.
• The project is independent and not affiliated with or endorsed by Anthropic; Claude and Claude Code are Anthropic trademarks.