# Using Claude Code with OpenRouter via `claude-code-router` on Linux

> Source: <https://dev.to/mbuyco/using-claude-code-with-openrouter-via-claude-code-router-on-linux-1f8p>
> Published: 2026-06-29 08:13:52+00:00

Claude Code is Anthropic's powerful command-line coding assistant that brings AI directly into your terminal, while OpenRouter is an API aggregator that provides access to multiple AI models through a unified interface. The `claude-code-router`

package bridges these two tools, allowing you to route Claude Code requests through OpenRouter instead of Anthropic's direct API.

This guide covers installation and configuration on Linux systems only.

Before beginning, ensure your system meets these requirements:

```
# Check Node.js version
node --version

# Check npm version
npm --version

# Check Git version
git --version
```

If Node.js is not installed or outdated, install it using your distribution's package manager or NodeSource:

```
# Ubuntu/Debian - Using NodeSource
curl -fsSL https://deb.nodesource.com/setup_20.x | sudo -E bash -
sudo apt-get install -y nodejs

# Fedora
sudo dnf install nodejs npm

# Arch Linux
sudo pacman -S nodejs npm
```

`claude-code-router`

The `claude-code-router`

package should be installed globally alongside Anthropic's official CLI tool.

```
npm install -g claude-code-router @anthropic-ai/claude-code
```

Create or edit your shell configuration file:

```
# For Bash
nano ~/.bashrc

# For Zsh
nano ~/.zshrc
```

Add the following lines:

```
# Claude Code Router Configuration
export OPENROUTER_API_KEY="your-openrouter-api-key-here"
export CLAUDE_CODE_ROUTER_MODEL="anthropic/claude-3.7-sonnet"
```

Apply the changes:

```
# For Bash
source ~/.bashrc

# For Zsh
source ~/.zshrc
```

Alternatively, create a configuration file in your home directory:

```
mkdir -p ~/.claude-code-router
nano ~/.claude-code-router/config.json
```

Add your configuration (updated with the correct Claude 3.7 model string):

```
{
  "apiKey": "your-openrouter-api-key-here",
  "model": "anthropic/claude-3.7-sonnet",
  "baseUrl": "https://openrouter.ai/api/v1",
  "maxTokens": 8192,
  "temperature": 0.7
}
```

If you prefer using an environment file over a JSON config, ensure you lock down the file permissions so other users on the Linux system cannot read your API key:

```
# Create a secure environment file
touch ~/.claude-code-router/.env
chmod 600 ~/.claude-code-router/.env
nano ~/.claude-code-router/.env
```

Add the key inside the file:

```
OPENROUTER_API_KEY=sk-or-v1-your-actual-key-here
```

One of the primary benefits of routing Claude Code through OpenRouter is natively leveraging **Prompt Caching** to heavily optimize your API costs.

When working in large codebases, Claude Code sends significant contextual data (file structures, previous messages, and active file contents) with every single request. OpenRouter automatically supports Anthropic's prompt caching functionality:

No extra configuration is required in `claude-code-router`

to enable this feature—OpenRouter handles the upstream caching headers automatically when translating requests to Anthropic's backend.

Configure Claude Code to point to the local proxy, then initialize the router:

```
export ANTHROPIC_BASE_URL="http://localhost:8080"
ccr code
```

If you want to quickly switch models without editing config files manually:

```
ccr ui
```

*Note: This command redirects to a local web UI to configure your preferred OpenRouter models.*

Check [OpenRouter's model list](https://openrouter.ai/models) for the most current routing options and model strings.

The `claude-code-router`

package provides a seamless way to use Claude Code through OpenRouter on Linux, giving you absolute flexibility in model selection while drastically cutting costs via prompt caching. By following this guide, you now have a secure, system-integrated proxy bridging Claude Code with OpenRouter's API infrastructure.

For additional help, consult the package's documentation on npm or GitHub, and periodically check OpenRouter's documentation for API changes or new model additions.