Privacy-First Coding: Using Claude Code with Self-Hosted Models via Open WebUI

Claude Code is a powerful tool for terminal-based development, but for many developers, sending sensitive codebase information directly to Anthropic’s servers is a dealbreaker. What if you could enjoy the seamless CLI experience of Claude Code while keeping your data strictly within your own infrastructure?

Because Open WebUI provides an API compatible with the Anthropic format, you can reroute Claude Code to your own self-hosted models. In this guide, I’ll show you how to configure Claude Code to use Open WebUI as your LLM provider, giving you the best of both worlds: advanced coding automation and total data sovereignty.

Get your Open WebUI API key

In Open WebUI, open the Settings page at the bottom left, then Account and finally scroll down to create or reveal your API key.

Install Claude Code

curl -fsSL https://claude.ai/install.sh | bash

Configure Claude Code

nano ~/.claude/settings.json

{
  "theme": "auto",
  "env": {
    "ANTHROPIC_BASE_URL": "https://your.openwebui.instance/api",
    "ANTHROPIC_AUTH_TOKEN": "sk-youropenwebuitoken"
  },
  "model": "gemma4:26b"
}

Important notes:

• the ANTHROPIC_BASE_URL must exactly match the URL of your instance, and end with /api without trailing slash /
• the model must exactly match the name of a reasoning model available on your Open WebUI instance

Test

You can now test your claude setup.

References

• https://docs.openwebui.com/reference/api-endpoints/#-anthropic-messages-api
• https://github.com/shanraisshan/claude-code-best-practice/blob/main/best-practice/claude-settings.md#model-configuration

This blog post is licensed under CC BY-SA 4.0