Products

Real products. Open them yourself.

No invented client logos, no fake testimonials. Everything here is something we built, and most of it you can try right now. It is the best evidence of how we will build yours.

Early access · Available now

Aura Browser

A privacy-first, agentic browser with three moods: Frost, Noir and Aurora. Built by our founder in seven months, now on Windows, macOS and Linux.

What’s inside

  • 51+ features: spaces, split view, tab groups, notes, reading list and more
  • Tor mode, DNS-over-HTTPS and fingerprint protection
  • 108K+ ad-block rules, running locally on your device
  • Built-in AI chat, password manager and breach monitor
  • Zero analytics, zero telemetry
What it proves for you: we can ship desktop-grade software on three platforms and do privacy properly, with the same discipline we would bring to your product.
ChromiumDesktopPrivacyAICross-platform
Download Aura ↗ Windows 10/11 · macOS · Linux
Live now

Connecto

Private chat without data mining. Create a room, share the code, and talk. Rebuilt from the ground up, with admin controls, stickers and GIFs, and file sharing that never touches a server.

What’s inside

  • Real-time chat with an instant emoji picker, original stickers and GIFs
  • Private rooms behind an 8-character code, with invite links
  • Room admins: the creator can remove people and keep a ban list
  • Photos, videos and files sent straight between phones, verified on arrival, never stored on a server
  • Google sign-in, light and dark themes, built phone-first
What it proves for you: real-time architecture, roles and permissions, peer-to-peer file transfer and security rules, designed, built and tested end to end by our team.
JavaScriptFirebaseWebRTCAuthSecurity rules
Live demo · Open source

Aizy

An AI-assisted store: shoppers get a fashion stylist that suggests outfits for any occasion, sellers get sales forecasts. A full-stack platform with a storefront, an Express API and a Python machine-learning service. Try the demo right on this page.

What’s inside

  • Next.js 15, React 19 and Tailwind frontend
  • Express.js backend with a dual-database design: MongoDB for products, MySQL with Prisma for users
  • Python service with FastAPI, scikit-learn and pandas for analytics and ML
  • Hosted demo mode that runs the whole store in your browser, with 50 sample products and no payments
What it proves for you: we can design and connect multi-service systems with AI at the core, not a chatbot bolted on afterwards.
Next.jsExpressFastAPIscikit-learnMongoDBMySQL
Free · Open source

Sceneseek

Video search for your own footage. Describe a moment and jump to the exact second. The AI runs on your computer, so nothing is uploaded, and every match wears a colour-coded percentage.

What’s inside

  • Search by words, by voice, or by picture, with dog + snow and beach -night power search
  • CLIP image and text models running locally with ONNX Runtime, no GPU needed
  • A match curve and film strip for every video, and topics found automatically
  • A reel that exports CMX3600 edit lists for Premiere Pro, DaVinci Resolve and Final Cut
  • One-click Windows installer, MIT licensed
What it proves for you: we can ship on-device AI as a polished desktop product, from the model and indexing pipeline to the interface, installer and website.
PythonFastAPICLIPONNX RuntimeSQLitePyInstaller
Try it here

Use our products without leaving this page.

Pick a demo and it loads below. They are sandboxes: nothing you type is sent to us, and no payments exist. Connecto needs real accounts, so try it on its own site.

Aizy store

Browse 50 products, ask the AI stylist what to wear, or sign in as a seller to see sales forecasts. Demo accounts are one click.

Runs entirely in your browser. Sample data only. Open full screen ↗
More we’ve built

Smaller projects, same craft.

Vireo answering a question about an imported PDF, with page citationsVireo model manager listing downloadable offline models
Android · Edge AI · Download

Vireo

An offline study assistant that runs a language model entirely on the phone: no internet, no cloud API. Chat with it or ask questions about your own PDFs and get answers with citations. A 15.5 MB app, with a thermal governor that keeps the phone cool.

Kotlinllama.cppNDKRAG

Android 9+ · 64-bit ARM · 15.5 MB · v0.1.0. You download a model inside the app (about 400 MB to start). Install steps and checksum.

The Context Compression Engine demo showing 853 tokens reduced to 436 for a question AI · Developer tool · Open source · v0.1

Context Compression Engine

A memory layer for AI assistants. It splits a long conversation into topics, keeps notes in hot, warm and cold tiers, and hands the model only the context a question needs. In our tests on local models it found facts said long ago with about 90% fewer tokens than resending the chat, as reliably as ordinary search; the benchmark page shows where it does not beat it. Works as a Python library, an MCP server, a REST API and a JavaScript client.

pip install context-compression-engine
npm install cce-client
PythonMCPEmbeddingsSQLiteNLP

Try the live demoPyPI ↗npm ↗GitHub ↗

On our roadmap

A relay server so Connecto file sharing works on every network · Aura updates and new features. We build in public, and this page stays honest about what is live and what is coming.

Start a project

Want something like this? Let’s make it yours.

Tell us about your idea. We will tell you honestly how we would build it.

Start a project ↗