Kimchi

Configure your AI coding tools to use Kimchi's open-source models in seconds.

---

What is Kimchi?

Kimchi is a CLI tool that configures your favorite AI coding assistants to use open-source models hosted by Kimchi:

Model	Role	Best For	Context	Output
kimi-k2.5	Primary model	Reasoning, planning, code generation, and image processing	262K tokens	32K tokens
nemotron-3-super-fp4	Coding subagent	Writing, refactoring, and debugging code	1M tokens	256K tokens
minimax-m2.7	Secondary subagent	Code generation and debugging (available across all tools)	196.6K tokens	32K tokens
nemotron-3-super-fp4	Reasoning model	High-performance reasoning for complex tasks	1M tokens	256K tokens

No API keys from Anthropic or OpenAI needed — just your Kimchi API key.

---

Quick Start

One-Line Install

curl -fsSL https://github.com/castai/kimchi/releases/latest/download/install.sh | bash

This downloads and installs Kimchi. Run kimchi to launch the Kimchi coding harness, or kimchi setup to configure another AI tool.

Manual Install

Download the latest release for your platform:

Platform	Architecture	Download
macOS	Intel	kimchi_darwin_amd64.tar.gz
macOS	Apple Silicon	kimchi_darwin_arm64.tar.gz
Linux	x86_64	kimchi_linux_amd64.tar.gz
Linux	ARM64	kimchi_linux_arm64.tar.gz

# Download and extract
curl -fsSL https://github.com/castai/kimchi/releases/latest/download/kimchi_linux_amd64.tar.gz | tar xzf -

# Make executable and move to PATH
chmod +x kimchi
sudo mv kimchi /usr/local/bin/

---

Getting Started

1. Get Your API Key

1. Go to app.kimchi.dev
2. Create and copy your API key

2. Run Kimchi

Running kimchi launches the Kimchi coding harness — a coding agent pre-wired to Kimchi's models. The harness binary is installed and kept up to date automatically alongside this CLI.

To configure another AI tool to use Kimchi instead, run:

kimchi setup

The interactive wizard will guide you through:

1. Auth — Enter and validate your Kimchi API key
2. Detect Tools — Automatically finds installed AI tools
3. Select Tools — Choose which tools to configure
4. Scope — Global (all projects) or project-specific
5. GSD Setup — Optional: Install Goal-Driven Development agents
6. Configure — Writes configuration files
7. Done — Ready to code!

---

Supported Tools

Tool	Description	Config File
OpenCode	Agentic coding CLI	~/.config/opencode/opencode.json
Claude Code	Anthropic's coding agent	~/.claude/settings.json
Codex	OpenAI's coding CLI	~/.codex/.env
Cursor	AI-powered code editor	state.vscdb (SQLite)
Windsurf	AI-powered IDE	globalStorage/storage.json
Zed	High-performance editor	~/.zed/settings.json
Cline	VS Code extension	~/.cline/data/globalState.json
OpenClaw	AI agent framework	~/.openclaw/openclaw.json
Generic	Environment variables	Prints to stdout

---

Usage

# Launch the Kimchi coding harness
kimchi

# Run the interactive wizard to configure other AI tools
kimchi setup

# Show version
kimchi version

# Show kimchi CLI help
kimchi --help

# Show coding harness help
kimchi --harness-help

# Update to the latest version
kimchi update

# Enable debug output
kimchi --debug

# Generate shell completion
kimchi completion bash > /etc/bash_completion.d/kimchi
kimchi completion zsh > "${fpath[1]}/_kimchi"
kimchi completion fish > ~/.config/fish/completions/kimchi.fish

---

How It Works

Kimchi configures each tool to use Kimchi's inference endpoint:

Your AI Tool ──► Kimchi Config ──► Kimchi Endpoint ──► Open-Source Models
                                        │
                                        ▼
                               https://llm.kimchi.dev

Configuration Example (OpenCode):

{
  "model": "kimchi/kimi-k2.5",
  "provider": {
    "kimchi": {
      "name": "Kimchi",
      "options": {
        "baseURL": "https://llm.kimchi.dev/openai/v1",
        "apiKey": "your-api-key"
      },
      "models": {
        "kimi-k2.5": { "reasoning": true },
        "nemotron-3-super-fp4": { "reasoning": true },
        "minimax-m2.7": { "reasoning": true }
      }
    }
  }
}

---

FAQ

Will this break my existing config?

No. Kimchi preserves your existing tool configurations and only adds its provider settings.

Can I switch back?

Yes. Simply remove the kimchi provider from your tool's config file, or re-run the tool's original setup.

Where is my API key stored?

• Config file: ~/.config/kimchi/config.json (permissions: 600)
• Environment variable: KIMCHI_API_KEY

How does telemetry work?

Kimchi collects anonymous usage data to help improve the tool, i.e., the CLI version running and the fact that the tool has been used. You can disable telemetry at any time:

kimchi config telemetry off         # disable via CLI
export KIMCHI_TELEMETRY=false       # disable via env var

To check your current telemetry status: kimchi config telemetry

---

Development

# Build
make build

# Test
make test

# Run locally
go run .

# Lint
golangci-lint run ./...

---

License

Copyright © Cast AI. All rights reserved.