Custom MCP servers
Five Model Context Protocol servers I built to wire Claude directly into my toolchain — analytics, email, deploys, search data, form leads. The agent doesn't just advise; it acts across the whole stack.
I'm Dominick Caruso — founder of Caruso Tech Services and a Director of IT in the field. I design, build, and ship AI-native products, client experiences, and tools, with AI in the loop from the first prompt to production.
I work alongside agents the way most engineers work alongside a framework. Here's the machinery behind the projects below.
Five Model Context Protocol servers I built to wire Claude directly into my toolchain — analytics, email, deploys, search data, form leads. The agent doesn't just advise; it acts across the whole stack.
An orchestrated chain of specialized agents — evaluate, build, review, security-gate, enhance — that audits a website and rebuilds it end-to-end, then grades its own output before it ships.
Local operations decks that pull live data — analytics, uptime, finances, search — into a single view I can query in plain language. One front door to everything I run.
Every product and client site here was designed, built, and shipped with AI in the loop. Not a bolt-on feature — the way the work gets made.
The day job, the agency, the products I'm chasing, and the things I build for fun. Each one's a filterable library — search, filter by status, or sort to see just what you want.
The day job — enterprise IT, automation & AI
Leading IT and data strategy for an environmental-services company through a company-wide transformation — modernization, automation, and analytics at operating scale.
Led company-wide AI adoption — team to enterprise — including metered-API workloads and scheduled “morning brief” agents so a lean team can offload routine work.
Designed an internal ops platform — workflow automation, uptime monitoring, AI-triaged email ticketing, self-service BI — with an AI reasoning layer on top.
Reverse-engineered a 400+ table legacy platform into a searchable ERD web app — ~493 entities, 157 inferred relationships — so an undocumented system finally became navigable.
A tool to design, document, and diff security roles with per-grant rationale, exporting an auditor-ready PDF. Read-only against the source by design.
A single front-door portal indexing every custom utility, query, and screen with lifecycle status and ⌘K search — a lightweight internal developer portal.
Authored a full methodology and reusable app shell for safely extending a closed industry ERP: naming standards, per-run audit and one-click reversal, versioning, UX patterns.
Wizard-driven importer that provisions new customers, sites, and services from a slim CSV in one reversible transaction — hours of manual entry gone per batch.
One-screen green/amber/red board for backups, scheduled jobs, sessions, locks, licenses, and disk — a five-second “is it OK?” without opening a SQL console.
Architected a near-real-time sync from closed legacy databases into a cloud Postgres mirror, so a customer-facing app can read live data without touching the system of record.
Lets dispatch slide all scheduled service calls forward or back by N days for selected routes in one reversible transaction.
Rebuilt a multi-entity financial close — automating the journal entries and reconciliations that used to eat days of manual work every single month.
A tiered AR toolset: a supervisor collections dashboard, an agent daily workspace with promise-to-pay tracking, and an executive cash-flow / DSO view.
Finds customers with email on file but no autopay and pushes that live cohort into the announcement system — modernizing payment methods and cutting paper and AR friction.
A Python tool that turns raw Word docs into a consistent, brand-styled HTML → print-ready PDF house format, so every policy and guide ships looking like one system.
Evaluated GPS and telematics for a large field fleet and modeled the operational upside — turning day-to-day operations into measurable, decision-ready data.
Full rebuild of a company's public site from WordPress to Next.js: 36 pages, an interactive facilities map, scroll animations, form autosave, privacy-first analytics, an AI chatbot scaffold.

By day I'm the Director of Information Technology for an environmental-services company, steering data and automation strategy through a company-wide transformation. By night — and most weekends — I run Caruso Tech Services, my own agency doing managed IT, cybersecurity, and AI-driven web work.
The thread between them is the same thing I've done since I was a kid: I build. Client sites, SaaS products, internal tooling, dashboards, even a Steam game — and lately, the AI machinery that lets one person ship like a team. I write the code, but I also build the agents that help write it.
I care about work that's fast, honest, and finished. If it has data, I'll build it a dashboard. If it's repetitive, I'll teach an agent to do it. If it's a good idea, I'll ship a prototype before I talk myself out of it.
Let's build something
I'm always up for interesting problems — client work, product collaborations, or just comparing notes on building with AI.