GitHub radar

Fullstack Agent: Voice, Memory and Face for Claude

A setup wizard for installing Claude Code’s persistent memory, backtalk (voice response), ai-visualizer (full screen animations) and barehands (hand gesture) to create a personal AI assistant.

01jaredrhod/fullstack-agent 349Batchfile

fullstack-agent is a setup wizard to install Claude Code’s personal AI assistant from 4 different open source projects: ai-memory-vault (persistent memory, stores all of Claude’s knowledge about you and your work in unlimited size text files), backtalk (voice response, press button, say something, hear response within ~1 second), ai-visualizer (full screen animations that change based on listen, think, speak), barehands (hand gesture control, works with a webcam without a headset). Everything is optional and can be installed individually. The setup wizard will ask you what you need. If something goes wrong, just tell Claude what happened and he will fix it for you with the included troubleshooting manuals. Updates are done with git pull and do not affect your personal information or agent identity. Available for macOS (Apple), Linux and Windows. On macOS, Apple Command Line Tools are included, macOS will prompt you to install them if you run it for the first time.

Why a vibe-coder should care

Transform Claude Code from a text editor to a voice controlled, persistent memory assistant perfect for long coding sessions. The individual parts are developed independently and constantly improved. Your personal data is stored separately and won’t be affected by updates.

How to install

Copy this and send it to your agent — Claude Code, Codex, any of them:

Install fullstack-agent from here: https://github.com/jaredrhod/fullstack-agent — run the README command for my OS (Mac/Linux/Windows), go through the setup wizard, enable memory and voice, then show me that the agent can hear me and reply out loud. Ask me if you need any keys or permissions.

Runs on a regular laptop — Mac, Linux, or Windows. The voice module downloads ~1 GB of local models on first launch.

Open on GitHub