Claude Code Kimi Groq — Development skill for Claude Code

Use Kimi K2 on Claude Code through Groq

This project builds upon the original repository at https://github.com/fakerybakery/claude-code-kimi-k2 (formerly `claude-code-kimi-groq`), enhancing it with additional features and improvements.

A proxy server that enables Claude Code to use Kimi K2 (Moonshot AI's model) through Groq's API by converting between Anthropic and OpenAI API formats.

Quick Start

Installation

# Install dependencies
pip install -e .

# Or with uv
uv run proxy.py

Running the Proxy

# Using uv (recommended)
uv run proxy.py

# Using python
python proxy.py

# Or using the CLI (after installation)
proxy

Configuration

Set environment variables before running Claude Code:

export ANTHROPIC_BASE_URL=http://localhost:7187
export ANTHROPIC_API_KEY=NOT_NEEDED
# Ensure GROQ_API_KEY is set in your environment or .env file

Then run Claude Code:

claude

Alternate Run Methods

# Using the shell scripts
./run.sh
# or
./run-claude.sh  # sets environment variables automatically

Architecture

The proxy handles:

• Message Format Conversion: Translates between Anthropic and OpenAI message formats
• Tool Handling: Converts tool schemas and function calls between API formats
• Response Mapping: Returns responses in Anthropic's expected format with proper message IDs and token usage

The server runs on port 7187 by default and uses the Kimi K2 model through Groq's API.

If you use this:

If you use this, I'd love to hear about your experience with Kimi K2 and how it compared with Claude! Please open an Issue to share your experience.

Acknowledgements

Inspired by [claude-code-proxy](https://github.com/1rgs/claude-code-proxy)

License

MIT