free-claude-code

Product Introduction

1. Definition: Free-Claude-Code (FCC) is an open-source, local proxy server and gateway application that provides free, seamless access to AI-powered coding assistants like Claude Code, Codex, and Pi. It operates as a middleware layer, routing requests from these agents to a wide array of free, paid, or local large language models (LLMs).
2. Core Value Proposition: It solves the critical problem of expensive or restricted access to advanced AI coding tools by enabling developers, students, and hobbyists to integrate powerful, multi-model AI coding assistance directly into their terminal, IDE, or messaging apps without vendor lock-in or high subscription costs.

Main Features

1. Multi-Agent, Multi-Provider Proxy: FCC acts as a unified gateway for multiple coding agents (Claude Code, Codex, Pi) to over 31 different AI model providers. It translates the native API calls from each agent into the correct format for the configured backend provider, whether it's a cloud service like NVIDIA NIM, OpenRouter, or Google Vertex AI, or a local instance like Ollama or LM Studio. This is achieved through a custom-built, provider-agnostic routing layer.
2. Centralized Admin UI for Model Management: The tool includes a local web-based Admin UI (default: http://127.0.0.1:8082/admin) that serves as a central control panel. Users can validate API keys, select from a dynamically populated dropdown of available models per provider, configure model-tier routing (e.g., route Claude Code's "Opus" tier to a specific model), and apply settings without manually editing configuration files.
3. Native IDE and Client Integration: FCC is designed for deep integration with existing developer workflows. It provides specific configuration guides for embedding Claude Code into VS Code and JetBrains IDEs, and for connecting the standalone Codex App. It generates a model catalog JSON file that allows native model pickers within Claude Code and Codex to display and select from all FCC-configured models seamlessly.
4. Advanced Routing and Reasoning Control: Beyond simple proxying, FCC offers granular control. Users can define specific models for different Claude Code tiers (Fable, Opus, Sonnet, Haiku), with MODEL acting as the global fallback. It also provides reasoning effort control (Off, Low, Medium, High, etc.), allowing users to override or inherit the reasoning level requested by the client to optimize for cost or performance per provider.
5. Optional Messaging Platform Bots with Voice: FCC can be extended with optional Discord and Telegram bot integrations. These bots allow users to interact with coding agents via chat, including support for voice-note transcription using either local Whisper models (CPU/CUDA) or NVIDIA NIM's speech-to-text API, enabling voice-command coding assistance.

Problems Solved

1. Pain Point: High cost and access barriers to proprietary AI coding assistants. Many advanced models like Claude Code or Codex require paid subscriptions, while free tiers are often limited. FCC democratizes access by routing to free-tier providers, open-source models, and local deployments.
2. Target Audience: The primary users are cost-conscious developers, indie hackers, students learning to code, and hobbyists experimenting with AI. It also serves professionals who need to use specific, non-standard models for coding tasks or who require a unified interface to switch between multiple AI providers for redundancy or benchmarking.
3. Use Cases: Essential for developers who want to use Claude Code's interface with a powerful, free model like Google's Gemini; for teams standardizing on a local, private model like Qwen Coder via Ollama; for users who want to code via voice commands on Telegram; or for anyone needing to bypass regional restrictions on certain AI services by using an alternative provider gateway.

Unique Advantages

1. Differentiation: Unlike single-provider wrappers or paid SaaS platforms, FCC is a free, open-source, and agnostic orchestration layer. It doesn't replace Claude Code or Codex; it enhances them by decoupling the client from the model provider. Competitors are typically closed-source, tied to one model, or lack the deep IDE integration and granular routing controls FCC offers.
2. Key Innovation: The core innovation is its provider abstraction and model catalog generation. It seamlessly makes dozens of disparate AI APIs (OpenAI-compatible, Anthropic-compatible, proprietary) appear as a single, unified "FCC" provider to the end-user's coding agent. This, combined with the tier-based routing and local Admin UI, creates a plug-and-play experience for advanced AI coding toolchain customization that is otherwise highly technical.

Frequently Asked Questions (FAQ)

1. Is Free-Claude-Code legal and safe to use? Yes, FCC is an open-source proxy tool (MIT licensed) that facilitates connections to legitimate AI model providers. It requires your own valid API keys for services like NVIDIA NIM, OpenRouter, or Google AI Studio. It does not pirate or crack any software; it simply reroutes traffic from official coding agent clients to your chosen backend.
2. Does using Free-Claude-Code violate Claude Code's or Codex's Terms of Service? FCC operates by configuring the official clients to connect to a user-controlled endpoint. Users are responsible for ensuring their use of the underlying AI models complies with the respective provider's (e.g., OpenAI, Anthropic, Google) and the client's Terms of Service. The tool itself is a configuration middleware.
3. What are the best free providers to use with Free-Claude-Code for coding tasks? For free coding assistance, providers like Google AI Studio (Gemini models), OpenRouter's free tier, DeepSeek, and locally run Ollama models (e.g., qwen2.5-coder) are popular choices. The "best" model depends on the specific coding language and task, requiring experimentation via the FCC Admin UI.
4. Can I run Free-Claude-Code and use local models without an internet connection? Yes, absolutely. Once FCC is installed and configured to use a local provider like Ollama, LM Studio, or llama.cpp, you can run your coding agents completely offline. The AI inference happens on your local machine, and FCC routes the requests locally.
5. How do I update Free-Claude-Code to the latest version? Updating is simple and uses the same installer script. Re-running the installation command (curl -fsSL ... | sh for macOS/Linux or the PowerShell command for Windows) will automatically update the FCC server, launchers, and dependencies to the latest release.