Claude code with Private AI

Tue, 05 May 2026 07:15:00 +0000

In this article we are going to see how to use Claude code with our own private models Gemma4, Qwen3, GPT OSS 120 or uncensored ones.

Prerequisites

LM Studio installed wherever you want

Installation

Install Claude Code

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

Configuration

Create a new file under ~/.claude/lmstudio.private-ai-server.json and add the following content:

{
  "env": {
    "ANTHROPIC_BASE_URL": "http://127.0.0.1:1234/",
    "ANTHROPIC_AUTH_TOKEN": "dummy",
    "API_TIMEOUT _MS": "3000000",
    "CLAUDE_CODE_DISABLE_NONESSENTIAL_TRAFFIC": 1,
    "ANTHROPIC_MODEL": "default_model",
    "ANTHROPIC_SMALL_FAST_MODEL": "default_model",
    "ANTHROPIC_DEFAULT_SONNET_MODEL": "default_model",
    "ANTHROPIC_DEFAULT_OPUS_MODEL": "default_model",
    "ANTHROPIC_DEFAULT_HAIKU_MODEL": "default_model"
  }
}

Test

Load a huge thinking Open Source model in LM Studio and set the context to the maximum limit. Then run the following command in your app repository claude --settings ~/.claude/lmstudio.private-ai-server.json. And finally select default_model using /model command after claude has started.

Claude on Lean Deep Tech blog

Claude code with Private AI

Prerequisites

Installation

Test