Platforms we have built for clients, the products we run ourselves, and the tools we make to build both faster.
Systems built for and owned by the businesses running them. These are private deployments, so there is nothing to click through to — what follows is what they do and what they are built on.
Veterinary clinic management platform
One platform covering the whole clinic: an admin console for staff, a public-facing web app for pet owners, a mobile app, and a shared API behind both. Appointments, patient records, and clinic operations run off a single schema rather than three systems that disagree with each other.
Multi-tenant booking and lead capture over WhatsApp
Customers message a business on WhatsApp — by text or by voice note — and a language model collects the booking details, confirms them, and drops every completed request into a staff dashboard. Voice notes are handled trilingually: callers mix Lebanese Arabic with English and French in a single sentence, and the bot understands them natively while preserving their exact words for staff to replay. Built multi-tenant from the start; the first operator runs a gym and physiotherapy practice across three branches.
Booking, packages, and session accounting for a sports venue
Members book courts in variable durations — one hour, ninety minutes, two hours, four — and prepaid session packages are debited by the actual time booked rather than a flat slot. Session deduction and refunds run as thread-safe database routines with row locking, so two people booking the same slot cannot both spend the same session. Admin tooling covers manual bookings, package management, and depletion handling.
What we build for ourselves, on our own budget and our own deadlines. It is also where we try things before we put them in front of a client.
Sticker and GIF sharing, built as a product
Our own app rather than client work. People share posts — images, video, animations — and turn them into stickers, including packs exported straight into WhatsApp. One React Native codebase serves both the mobile app and the web app, and the backend is fully serverless: single-table DynamoDB, a typed GraphQL contract, and passwordless auth because nobody wants another password.
Agents, plugins, and command-line tooling we wrote because the work needed them. Several are open source; all of them are in daily use on client projects, which is the only test of a developer tool that matters.
Design-intelligent MCP server
201 tools across 17 categories that let an AI coding agent work inside Figma properly — with the 8px grid, type scales, semantic colours, accessibility checks, and component defaults built into every tool rather than left to the model to remember.
Operator system for design, product, and brand work
Agents, skills, hooks, rules, and commands kept as one source of truth in Markdown, with adapters for six different AI coding harnesses. Write the way your team works once; every tool reads the same definitions. MIT licensed and meant to be edited.
Social campaign operator
Campaign strategy, platform-native content, engagement and comment playbooks, and insight-driven listening for X, LinkedIn, Discord, and Telegram — with output that does not read like a bot wrote it, which is the entire difficulty of the problem.
Structure for creative AI work
Ask a model to build a brand and you get competent, generic output — no taste, no constraints, no memory of what came before. This is the missing structure: 25 named agents, 5 overnight pipelines, and 8 connected workflows carrying brand strategy, design systems, product thinking, and growth mechanics.
Corporate photography, generated
A desktop application for producing business portraits at volume — headshots, environmental portraits, team sets, conference and brand photography. Save each person once, choose the looks, and a background queue renders the whole combination. One sitting, every portrait the company needs.
Hands on the internet for an AI assistant
A small, fast command-line agent written in Go that gives an assistant the ability to actually go and do things online rather than describe them.
Most of the work above started as a workflow somebody was doing by hand. If you have one of those, we can tell you quickly whether it is worth automating and what it would cost.