Free Claude Code

Free Claude Code is a local proxy gateway that lets you run Claude Code through 17+ alternative model providers — NVIDIA NIM, OpenRouter, Google Gemini, DeepSeek, Mistral, Groq, Cerebras, Fireworks AI, Ollama, llama.cpp, LM Studio, and more — for free or at a fraction of the standard Claude API cost.

What is Free Claude Code?

Free Claude Code (sometimes referenced as fcc) is an open-source proxy server that sits between Claude Code and Anthropic's API. Instead of routing your Claude Code calls to api.anthropic.com, it routes them to any of 17+ supported providers based on your routing rules. The model picker, tool use, streaming, and reasoning blocks all keep working — Claude Code never sees the difference.

It is built and maintained by Alishahryar1 under an MIT license, with 34k+ stars and 5.2k+ forks at the time of this writing. The implementation is Python 3.14 with FastAPI/Uvicorn, tested with Pytest, formatted with Ruff, type-checked with Ty, and logged through Loguru.

What it does

• Drop-in proxy for Claude Code's Anthropic API calls — no Claude Code patches required, just environment variables.
• 17 provider backends — NVIDIA NIM, OpenRouter, Google AI Studio (Gemini), DeepSeek, Mistral La Plateforme, Mistral Codestral, OpenCode Zen, OpenCode Go, Wafer, Kimi, Cerebras Inference, Groq, Fireworks AI, Z.ai, LM Studio, llama.cpp, and Ollama.
• Per-model routing — send Opus-class, Sonnet-class, Haiku-class, and fallback traffic to different providers from a single config.
• Native /model picker support through the proxy's /v1/models endpoint.
• Optional Discord or Telegram bot wrapper for remote coding sessions.
• Optional voice-note transcription through local Whisper or NVIDIA NIM.
• Local Admin UI at /admin to edit settings, validate changes, and check provider health.

How it works

You install the proxy locally with one shell command, start it with fcc-server, and either run Claude Code via the wrapper fcc-claude (which sets the right env vars for you) or point your IDE's Claude Code extension at the proxy directly. The proxy listens on http://127.0.0.1:8082 by default, exposes the same endpoints as Anthropic's API, and translates requests on the fly to whichever backend you have configured for the requested model.

Routing rules live in the admin UI. You can assign different providers per Claude model tier — for example, Opus-class calls to OpenRouter, Sonnet-class to Gemini, Haiku-class to Groq — set fallback chains, and toggle features like streaming or tool use per provider. Configuration changes hot-reload without restarting the proxy.

Who it is for

• Developers who want to run Claude Code without an Anthropic subscription, using free tiers from NVIDIA NIM, OpenRouter, Google AI Studio, or DeepSeek.
• Teams that want to mix providers per workload — premium models on critical tasks, cheap or free models on routine work.
• Privacy-conscious users who want to run Claude Code entirely against local models via Ollama, LM Studio, or llama.cpp.
• Power users running Discord or Telegram bot integrations for remote coding sessions from a phone.
• Cost-sensitive teams in regions where Anthropic billing is impractical.

What you can build

• A fully free local coding agent using only NVIDIA NIM or OpenRouter's free tier.
• A multi-tier coding setup that routes by model class for cost optimisation.
• An air-gapped Claude Code experience using only Ollama and llama.cpp models.
• A team Discord bot that runs Claude Code on demand without per-seat licenses.
• Voice-to-code workflows using Whisper or NVIDIA NIM transcription as the input layer.
• A multi-provider failover setup that automatically routes around rate limits.

Why it matters

Claude Code is the best AI coding agent on the market, but its cost can be a real constraint for solo developers, students, and teams in cost-sensitive regions. Free Claude Code keeps the Claude Code experience intact — same CLI, same IDE extensions, same model picker, same tool use — while opening up access to free and lower-cost model providers. For many users it is the difference between using Claude Code daily and using it occasionally.

1. Install

macOS / Linux:

curl -fsSL "https://github.com/Alishahryar1/free-claude-code/blob/main/scripts/install.sh?raw=1" | sh

Windows PowerShell:

irm "https://github.com/Alishahryar1/free-claude-code/blob/main/scripts/install.ps1?raw=1" | iex

2. Start the proxy

fcc-server

The proxy listens on http://127.0.0.1:8082 by default.

3. Configure providers

Open the local admin UI at http://127.0.0.1:8082/admin to add API keys, select providers, and configure per-model routing. Set fallbacks if you want the proxy to retry on a different backend when one fails.

4. Run Claude Code through the proxy

fcc-claude

The wrapper sets the right environment variables and launches Claude Code pointing at your local proxy.

5. Use with VS Code or JetBrains extensions

For the VS Code Claude Code extension, add to settings.json under claudeCode.environmentVariables:

• ANTHROPIC_BASE_URL: http://localhost:8082
• ANTHROPIC_AUTH_TOKEN: freecc
• CLAUDE_CODE_ENABLE_GATEWAY_MODEL_DISCOVERY: 1
• CLAUDE_CODE_AUTO_COMPACT_WINDOW: 190000

JetBrains ACP integration follows the same environment-variable pattern.