Skip to main content

​
Overview

OCA installs AI CLI tools via npm install -g during Step 7. All run through the glibc Node.js environment.
AI CLIs need API keys from their respective providers. OCA installs the binaries — you configure the keys after install.
Local LLM Support: Want to run models locally on your phone? See Local LLM Guide for node-llama-cpp and Ollama setup.

​
Qwen Code

Best choice for Android — lightweight and uses --ignore-scripts which skips native builds that fail on Android. Install: 
npm install -g @qwen-code/qwen-code@latest --ignore-scripts
Setup: 
qwen config set api-key YOUR_DASHSCOPE_KEY
qwen config set model qwen-coder-turbo
Start: 
qwen
--ignore-scripts skips node-pty and @lvce-editor/ripgrep native builds, which both fail on Android. Core CLI functionality is unaffected.

​
Claude Code

npm install -g @anthropic-ai/claude-code
claude config set apiKey YOUR_ANTHROPIC_KEY
claude
Get your key at console.anthropic.com.

​
Gemini CLI

npm install -g @google/gemini-cli
gemini  # prompts for Google account login on first run
Or with API key: 
export GEMINI_API_KEY=your_key
gemini
Get your key at aistudio.google.com.

​
Codex (OpenAI)

npm install -g @openai/codex
export OPENAI_API_KEY=your_key
codex

​
Using inside tmux

Always run AI CLIs inside tmux for long sessions: 
tmux new -s ai
qwen
# Detach: Ctrl+B, then D
# Reattach: tmux attach -t ai

​
Managing CLIs

# Check installed
npm list -g --depth=0

# Update all
npm update -g @qwen-code/qwen-code @anthropic-ai/claude-code @google/gemini-cli @openai/codex

# Or via OCA
oca --update