Use Claude Code with OpenAI, Azure, Databricks and More 🤝

A proxy server that mirrors Claude Code's interface to OpenAI models, Azure OpenAI deployments, and other compatible providers. 🌉

Clone the repository and install locally:

# Clone the repository
git clone https://github.com/ericmichael/claude-mirror.git
cd claude-mirror

# Run the install script (recommended)
./install.sh

The install script will:

1. Install the package using your current Python environment
2. Create a launcher script that works across Python environments
3. Offer to install the launcher globally to make it available from any terminal

You can also install manually if preferred:

# With uv (recommended)
uv pip install -e .

# Or with pip
pip install -e .

If you install manually, you may need to ensure your Python environment is activated when running commands, or use the Python module syntax:

python -m claude_mirror.cli

Once installed, run the interactive setup:

claude-mirror --setup

This will guide you through creating a configuration file with your API keys and model mappings. After setup, run:

claude-mirror

• OpenAI API key 🔑
• Optional: Azure OpenAI or Databricks endpoints

1. Install the package:

   # Clone the repository
   git clone https://github.com/ericmichael/claude-mirror.git
   cd claude-mirror
   
   # Run the install script
   ./install.sh

2. Run the interactive setup:

   claude-mirror --setup

   This will guide you through creating a configuration file at ~/.claude-mirror/config.yaml.

   Alternatively, you can manually create a configuration file:

   # Provider configuration
   providers:
     # OpenAI configuration
     openai:
       api_key: your_openai_api_key_here
     
     # Optional: Azure OpenAI configuration
     # azure:
     #   api_key: your_azure_api_key_here
     #   endpoint: your-instance.openai.azure.com
     #   api_version: 2023-05-15
     
     # Optional: Databricks configuration
     # databricks:
     #   token: your_databricks_token_here
     #   host: adb-12345678901234.12.azuredatabricks.net
   
   # Direct mapping from model categories to provider-specific models
   model_categories:
     # Claude models will map to these categories
     large: openai/gpt-4o          # Claude-3-Sonnet maps to this
     small: openai/gpt-4o-mini     # Claude-3-Haiku maps to this

3. Run Claude Mirror:

   claude-mirror

# Normal usage (no arguments needed)
claude-mirror

# Interactive setup to create/update config
claude-mirror --setup

# Debug mode with detailed logs
claude-mirror --debug  

# Use a specific config file
claude-mirror --config /path/to/my/config.yaml

1. Install Claude Code (if you haven't already):

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

2. The claude-mirror command handles everything:

   • Starts the proxy server
   • Configures the environment
   • Launches Claude Code connected to the proxy
   • Automatically shuts down the proxy when you exit Claude

3. That's it! Your Claude Code client will now use your configured models through the proxy. 🎯

The proxy follows strict rules with no fallbacks:

1. Category-based Mapping:

   • large and small categories in config.yaml MUST be defined
   • Claude Sonnet models → map to the "large" category
   • Claude Haiku models → map to the "small" category

2. Direct Model References:

   • Each category maps directly to a provider/model in the format: provider/model-name
   • For example: large: openai/gpt-4o means Claude-3-Sonnet requests will use OpenAI's GPT-4o
   • No default fallbacks or silent provider selection

3. Error Conditions:

   • Missing required categories results in clear errors
   • Invalid model format results in clear errors
   • Missing configuration values in config.yaml result in clear errors

The configuration is based on two simple concepts:

1. Providers: Where to get models from (OpenAI, Azure, etc.)
2. Model Categories: How Claude models map to your chosen provider models

The simplest setup uses OpenAI models:

providers:
  openai:
    api_key: your_openai_api_key_here
    # Optional: base_url for alternative OpenAI-compatible APIs
    # base_url: http://localhost:8000/v1

model_categories:
  large: openai/gpt-4o
  small: openai/gpt-4o-mini

You can use other OpenAI-compatible APIs by configuring a custom base URL:

providers:
  openai:
    api_key: your_api_key_here
    base_url: http://localhost:8000/v1  # Example for LocalAI

model_categories:
  large: openai/local-model-name
  small: openai/smaller-local-model

This works with services like LocalAI, LM Studio, or any other API that implements the OpenAI interface.

To use Azure OpenAI Service:

providers:
  azure:
    api_key: your_azure_api_key_here
    endpoint: your-instance.openai.azure.com
    api_version: 2023-05-15

model_categories:
  large: azure/my-gpt4-deployment-name
  small: azure/my-gpt35-deployment-name

To use Databricks:

providers:
  databricks:
    token: your_databricks_token_here
    host: adb-12345678901234.12.azuredatabricks.net

model_categories:
  large: databricks/databricks-claude-3-sonnet
  small: databricks/databricks-claude-3-haiku

You can also mix providers if needed, for example using OpenAI for one category and Databricks for another.

This proxy works by:

1. Loading configuration directly from config.yaml with strict validation (no defaults)
2. Mapping model names according to explicit rules in the configuration
3. Routing requests to the appropriate provider based on model prefix
4. Converting responses back to Anthropic format

The proxy maintains compatibility with Claude Code while ensuring strict configuration control using a single configuration file with no environment variables.

Contributions are welcome! Please feel free to submit a Pull Request. 🎁