Free Claude Code: How to Set It Up on Windows and Mac

By Shah Rukh, software developer · October 2, 2026 · 9 min read

Free Claude Code (FCC) is an open-source program that lets you use the Claude Code coding assistant with free AI models from other companies, so you don’t need a paid Claude plan to try it. This guide shows the full setup on Windows and Mac, and the precautions to take before you start.

Read this first. FCC does not give you Anthropic’s Claude models (Opus, Sonnet, Haiku) for free. You get the Claude Code app, but the answers come from another model, such as one on NVIDIA’s free tier. It is an independent project, not made or endorsed by Anthropic.

What Free Claude Code is

Claude Code is a coding assistant that runs in your terminal. It reads your project, writes and edits files, and runs commands. Normally it talks to Anthropic’s servers, which needs a paid plan or paid API credit.

FCC is a small server that runs on your own computer and sits in the middle. Claude Code sends its requests to FCC, and FCC passes them to a provider you choose. Its project file describes it as a “local proxy connecting coding agents to OpenAI-compatible AI providers”.

The project’s own page lists 59 providers, including NVIDIA NIM, OpenRouter, Groq and Google AI Studio, plus local options such as Ollama and LM Studio. It also works with other coding agents such as Codex, OpenCode, Cline and Aider. Free limits are set by each provider and can change at any time.

What you need

• A computer with Windows 10 or 11, or a Mac.
• An internet connection and about 15 minutes.
• A free NVIDIA account for the API key (or an account with another provider).

You don’t need to install Python yourself. The installer sets up a tool called uv, the exact Python version FCC needs, FCC itself, and the coding agents you choose.

Precautions before you install

1. Check the installer first. The install command downloads a script and runs it. The project links both scripts so you can read them: install.sh (Mac) and install.ps1 (Windows). Both also have a dry-run option that prints every step without doing anything. It’s shown in the steps below.
2. Only use the real repository. Copy commands from github.com/Alishahryar1/free-claude-code. A popular name attracts fake copies.
3. Your code leaves your computer. Whatever the assistant reads is sent to the provider you choose. Free tiers may store it or use it to improve their models. Don’t use a free cloud model on client work, company code or anything with passwords in it. For private code, use a local model (see below).
4. Keep your API keys secret. FCC saves its settings in a folder called .fcc in your user folder. Don’t share that folder, and don’t post screenshots of the Admin page with a key visible.
5. Back up your project. A coding agent edits files and runs commands. Use it inside one project folder, keep the project in Git or make a copy first, and read each permission prompt before you approve it.
6. Expect weaker results than real Claude. Free models make more mistakes on large tasks. Review what it writes.

How to install Free Claude Code on Windows

1. Click Start, type PowerShell and open Windows PowerShell. (Not Command Prompt.)
2. Optional but sensible: run a dry run first. It only prints what the installer would do.

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

3. Paste the install command and press Enter:

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

4. The installer asks which coding agents you want. Answer Y for Claude Code. You can answer N to the others.
5. It also asks about RTK, an optional add-on that trims long command output to save tokens. You can skip it for now.
6. Wait for it to finish, then close PowerShell and open it again so the new commands are found.

How to install Free Claude Code on Mac

1. Press Cmd + Space, type Terminal and press Enter.
2. Optional but sensible: run a dry run first.

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

3. Paste the install command and press Enter:

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

4. Answer Y for Claude Code when asked which coding agents to install. RTK is optional.
5. When it finishes, close Terminal and open it again.

The same command works on Linux.

Start Free Claude Code

• Windows: open Free Claude Code from the desktop or the Start menu. An icon appears in the system tray, near the clock.
• Mac: open Free Claude Code from the desktop or the Applications folder. An icon appears in the menu bar at the top.
• Linux: run fcc-server and keep that terminal open.

FCC opens its Admin page in your browser. Use the tray or menu-bar icon to open Admin again, restart FCC or quit it. FCC must stay running while you use the assistant.

Get a free NVIDIA NIM API key

The project uses NVIDIA NIM as its default provider.

1. Go to build.nvidia.com/settings/api-keys and sign in or create a free account.
2. Create an API key and copy it.
3. In the FCC Admin page, paste it into the field named NVIDIA_NIM_API_KEY.
4. Leave MODEL on the default, or search the dropdown and pick another model.
5. Click Apply.

Run Claude Code through FCC

1. Open PowerShell (Windows) or Terminal (Mac).
2. Go to your project folder. Change the path to your own:

   cd Documents/my-project

3. Start Claude Code through FCC:

   fcc-claude

4. Type a small request to test it, for example: explain what this project does.

Always start it with fcc-claude. The plain claude command goes to Anthropic and asks you to log in. Inside Claude Code, type /model to switch between the models you set up in FCC.

Lock down the local server

This is the step most guides skip. In the code we reviewed, the default Server Host is 0.0.0.0 on port 8082. That means the server accepts connections from other devices on the same network, not only from your own computer. On home Wi-Fi that is a small risk. On hostel, office or café Wi-Fi, someone else could use your API key through it.

Do one of these in the Admin page:

• Turn on Proxy Authentication, so every request needs a secret token. The project recommends this for protecting the local proxy.
• Or change Server Host to 127.0.0.1, so only your own computer can connect. FCC needs a restart after this change.

Other free providers and private local models

To add another provider, open its key page, paste the key into the matching field in Admin, pick a model in the MODEL dropdown and click Apply.

Fallback models: under Model Config you can add an ordered list of backup models. If the first provider fails or hits its limit, FCC tries the next one. One request can then use up quota at more than one provider.

Local models: with Ollama or LM Studio the model runs on your own computer, so your code is not sent to a cloud provider. You need a reasonably powerful machine, and the project advises choosing a model that supports tool use and has enough context for a coding agent.

ollama pull llama3.1
ollama serve

Paid accounts: FCC can also connect a ChatGPT or GitHub Copilot subscription. Read that service’s terms before you connect a paid account to any third-party tool. If you’re unsure, stay with normal API keys.

Update, check and uninstall

To uninstall, quit FCC and run the command for your system. It removes FCC, its launcher and the .fcc settings folder. It keeps uv, Python and your coding agents.

Windows PowerShell:

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

Mac and Linux:

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

Common problems

For anything else, run fcc-doctor and include its output when you report the problem on GitHub.

This guide is based on the project’s repository as it was on October 2, 2026. The project changes often, so if a step looks different, follow the README on GitHub. ToolsCloset is not connected to the project or to Anthropic. Claude and Claude Code are trademarks of Anthropic.

More AI tips

Related tools