MiMo Code Install Guide

Node.js (npm install only)

The official installer (Method B) downloads a self-contained binary and needs no Node.js. Node.js is only required for the npm install method (Method A); no specific version is declared by the package.

brew install node
node --version  # confirms Node.js is installed (npm route only)

# In WSL terminal:
sudo apt update && sudo apt install nodejs npm -y
node --version  # confirms Node.js is installed (npm route only)

curl -fsSL https://deb.nodesource.com/setup_lts.x | sudo -E bash -
sudo apt install -y nodejs
node --version  # confirms Node.js is installed (npm route only)

Install MiMo Code CLI

Two methods — pick one.

Method A: npm (recommended)

npm install -g @mimo-ai/cli

Method B: curl install script

curl -fsSL https://mimo.xiaomi.com/install | bash

⚠️ Security tip: Always review install scripts before piping to bash. The official install command is documented in the GitHub README.

Verify Installation

mimo --version

Expected output: a version number, for example 0.1.15.

Set Up Your Model Provider

The first launch walks you through provider setup (Xiaomi MiMo Platform OAuth login, Codex login, Claude Code import, or an API key). To configure an OpenAI-compatible endpoint yourself, save the config below as .mimocode/mimocode.jsonc in your project, or as ~/.config/mimocode/mimocode.jsonc for all projects. The apiKey is stored as plaintext, so keep the file private and never commit it. Verify with mimo models.

Option A: OpenRouter (flexible, multi-model)

{
  "$schema": "https://mimo.xiaomi.com/mimocode/config.json",
  "model": "custom/anthropic/claude-sonnet-4",
  "provider": {
    "custom": {
      "name": "OpenRouter",
      "npm": "@ai-sdk/openai-compatible",
      "only_configured_models": true,
      "models": { "anthropic/claude-sonnet-4": { "name": "Claude Sonnet 4" } },
      "options": { "baseURL": "https://openrouter.ai/api/v1", "apiKey": "your-openrouter-key" }
    }
  }
}

Option B: DeepSeek (affordable)

{
  "$schema": "https://mimo.xiaomi.com/mimocode/config.json",
  "model": "custom/deepseek-chat",
  "provider": {
    "custom": {
      "name": "DeepSeek",
      "npm": "@ai-sdk/openai-compatible",
      "only_configured_models": true,
      "models": { "deepseek-chat": { "name": "DeepSeek Chat" } },
      "options": { "baseURL": "https://api.deepseek.com", "apiKey": "your-deepseek-key" }
    }
  }
}

Option C: Local (private, free)

// Start Ollama first: ollama serve
// Then pull a model: ollama pull qwen2.5-coder:14b

{
  "$schema": "https://mimo.xiaomi.com/mimocode/config.json",
  "model": "custom/qwen2.5-coder:14b",
  "provider": {
    "custom": {
      "name": "Ollama",
      "npm": "@ai-sdk/openai-compatible",
      "only_configured_models": true,
      "models": { "qwen2.5-coder:14b": { "name": "Qwen 2.5 Coder 14B" } },
      "options": { "baseURL": "http://localhost:11434/v1", "apiKey": "ollama" }
    }
  }
}

Launch MiMo Code in Your Project

cd /path/to/your/project
mimo

On first run, MiMo Code will initialize its configuration. Run /init to set up the project context.

Optional: Security Hardening

Disable telemetry if you prefer privacy:

export MIMOCODE_ENABLE_ANALYSIS=false

Add the variable to your shell profile (~/.zshrc, ~/.bashrc) for persistence.