Shortcuts Toolbox has a Home

Shortcuts Toolbox is now a real, paid macOS app with a proper landing page. It's the spiritual successor to LaunchCuts and PushCut–dependency graphs for your Shortcuts library, scheduling, a REST API for triggering them, and an AI chat interface that works with Claude, Gemini, or GPT (your choice of provider). It also ships with a built-in MCP server, so you can expose your Shortcuts as tools to Claude, Cursor, or any MCP client–use them yourself, or share with the world.

Time Memos (Beta)

I'm beta testing Time Memos, a tiny iOS app for stamping a moment in time. Add an emoji, a label, optionally a location, and you've got a timestamped event you can search and filter later. Lock screen and Dynamic Island widgets, Siri Shortcuts integration, and CloudKit sync across devices. It's the kind of utility I always wanted but never found, so I built it. Working through TestFlight feedback before a public release.

recrd — The Centralized Contextual Brain for AI

The project I've been most heads-down on is recrd. It's an npm package and CLI for defining, managing, and querying structured data, with a built-in browser UI (recrd browse), an HTTP API, and a Node module API–four co-equal surfaces over the same core. Records live in your repo as JSON, validated against a JSON Schema you define.

The pitch: "so many systems are about the flow of data; recrd is about the stow and show of data." It's where your structured data comes to rest and be understood. Network devices, tasks, notes, bookmarks, Jira tickets–whatever you want to track, you define the shape, pipe the data in, and query it with MongoDB-style filters from the CLI, the browser, or an AI agent. Every entity can carry llm_instructions that teach an agent how to work with it, which is the real unlock: recrd becomes a shared substrate that any AI tool (Claude Code, Cursor, Copilot, your own scripts) can read and write against, and the value compounds as your structured context accumulates over time. I'm dogfooding it to track recrd's own roadmap.

ai-cli — My Personal AI Toolkit

I've been steadily growing ai-cli, a personal grab-bag of AI tools I use every day:

  • ai chat — a terminal ChatGPT/Claude client with history, resume, custom system prompts, and named sessions.
  • ai jared-server — wires GPT/Claude into iMessage via Jared, so my friends can text me and get an AI on the other end. I wrote about this idea years ago; it's great to finally have it running.
  • ai readwise — my AI-powered "newspaper." It pulls articles from Readwise Reader into a local SQLite store, runs them through an evaluation pipeline (triage, relevance, duplicate detection, tag suggestions, YouTube transcript analysis), and gives me an Angular web client where I can chat with an article, add tags, triage it, or kick it to "later." It's my morning reading ritual on autopilot.
  • ai claude-export — exports Claude Code conversations to markdown so I can archive and search them outside the app.

@prismcodebase — Slackbot at Work

At Prism I built @prismcodebase, a Slackbot the whole team can ping in any channel to have a full conversation with our codebase. Non-technical CS staff can start a debugging conversation that ends with a ticket or a PR, no engineer in the loop.

Under the hood it's an orchestrator over 23+ tools–select db tables, Jira, Confluence, GitHub, Perplexity, Claude Code sessions, and a whole slew more–with Claude Sonnet 4 / Opus 4.6 as the decision engine and the Vercel AI SDK driving tool selection and chaining (up to 30 steps). Heavy operations run as background jobs, per-thread context is preserved on disk, and a fast model posts instant acknowledgments so users aren't left wondering. Before the bot, engineers were spending hours every day helping CS with this stuff–it now handles ~90% of that work. We considered productizing it but decided to just write up the method and share it with the community; the link above is the full breakdown.

Mingo — Event Monitoring

I've been building Mingo with my friend Adrian–a service that scans social profiles for local events, extracts the event data with AI, and notifies you when something matches your interests. It scratches my own itch–I kept missing shows and pop-ups that were only announced on social–and it's turning into a real product. There's a business plan, a roadmap, and a real backend (Node/Express, Prisma, Cloudflare R2, Scrapfly for scraping, Claude + Gemini for extraction).

DJ Request

DJ Request, which I've been building with my friend Malcolm, is a song request app for DJs. Guests scan a QR code, browse the DJ's accepted request list, and submit songs from their phone; the DJ sees a live queue with real-time voting over WebSockets. Magic-link email auth, admin/DJ roles, Angular frontend, Node/Express backend with SQLite.

More Apple Tips

I've been steadily adding new tips to apple.lukezilioli.com. It started as a place to collect my Apple Wishlist posts, but it's slowly turning into a broader knowledge base for the little iOS/macOS tricks I keep rediscovering and want to remember.