A multi-agent AI assistant, a macOS MCP control center, a native video production tool, and a shelf of smaller experiments. Most projects below are public on GitHub; Studio is currently in public beta.
A model-selectable AI assistant with a ReAct coordinator, 84 Google Cloud tools, browser automation via Playwright, four-layer memory, scheduled workflows, and media generation. Runs on macOS and iOS with a Tauri + React desktop client and a native SwiftUI app.
A native macOS menu bar app for quick access to QMD memory search, indexing status, and collection management. Search your agent memory vault, trigger reindexing, and browse collections without leaving the menu bar.
A native macOS production workspace for AI-generated video and media. Contextual assistant actions, reviewable artifacts, and persistent .studio project packages instead of disposable chat threads. In private beta.
A model-agnostic HUD for the MacBook notch. Displays live context, tool status, and agent state without leaving the active workspace. Private, in development.
A native macOS control center for local MCP servers: inventory configs, inspect running processes, and surface issues.
A cowbell for your agent to ring.
External project I use regularly: unicode spinners, ASCII loaders, and CSS/SVG/Lottie animations for loading and generating states.
A minimal Electron desktop app for downloading YouTube videos as MP3 or MP4 files.
An open-source e-reader for experimenting with a quieter, more focused reading interface.
A retired Google Workspace MCP integration for querying Gmail, Drive, GA4, and Merchant Center through agent workflows. Functionality moved to the GWS CLI.
A multi-agent research experiment for literature workflows and structured text analysis.
An early multi-agent LLM project for testing coordination patterns and tool-using assistant behavior.
A fine-tuning experiment for Mistral models using QLoRA-style workflows.
