I've been using Claude Code daily, and my biggest frustration was babysitting it. You give it a task, wait for it to finish, then give it the next one. If it asks a permission question, you have to be at your keyboard to respond. You can't really walk away.

So I built qlaude. It's a CLI wrapper that adds two things to Claude Code:

1) A queue system — write prompts in a text file and qlaude feeds them to Claude one by one automatically.

2) Telegram integration — when Claude hits a selection prompt, qlaude sends it to your phone as inline buttons. Tap to respond, Claude continues.

A queue file looks like this:

  @model sonnet
  Set up Express + TypeScript project

  @new
  @model opus
  Implement CRUD API. Ask me which DB to use.

  @new
  @model sonnet
  Write integration tests

It works by spawning Claude Code in a PTY, maintaining a headless xterm screen buffer, and analyzing the rendered output to detect Claude's state. No API calls, no Claude Code modifications — just a wrapper around the terminal.

Free, open source, MIT licensed. Still in alpha. Would appreciate any feedback.