Skip to content
Foldox
Menu
FoldoxStart a project
Menu

An AI software lab.We design, build and run AI software.

Apps, agents, image and video pipelines, and assistants that answer from your data, from the first prototype to production.

Start a projectSee selected work

How we build

From an idea to shipped software.

  1. We learn the problem, your data and what a good result looks like.

    Example brief: the goal, what a good result looks like, and the data you already have, such as a help centre, order history and policy PDFs.

  2. We pick the models and the architecture, and agree scope and cost.

    Example architecture sketch: your data goes through retrieval to a model picked for each step, then to an agent that waits for a person to approve, then into your app.

  3. A working version early, tested on real inputs.

    Example wireframe of a first working screen, with one note: tried on real inputs.

  4. We launch it on the web or in the app stores.

    Lovino. AI image and video generation. Type a prompt, pick a model, get the result.

Why Foldox

Foldox is an AI software lab. We take an idea from the first conversation to software running in production, and we can stay on to run it after launch.

Every project starts with your problem, not a model. We learn the work, the data you already have and what a good result looks like. The model is picked last, for the task.

You see working software early. Something you can try on your own inputs, so decisions come from what it actually does rather than from a slide.

Assistants we build show their sources. When the answer isn’t in your documents, they say so instead of guessing.

Your data stays yours. We use it for your project only, never to train shared models.

And a person stays in the loop. Agents handle the routine steps and stop for approval before anything that needs judgement.

Ready to build yours?

Start a project

Services

What we build.Six kinds of AI software.

Pick one to see the kind of screen we build, with sample data.

Tell us what you’re building

01 / 06

Custom AI apps and features

New AI products, or AI features inside the product you already have.

  • From first idea to launch
  • Web apps with Next.js and TypeScript
  • Accounts and payments around the AI
Illustration: an existing accounts table with an AI summary panel added to it. Sample data.

Inside the work

A closer look.

Three patterns from our shipped work.

See selected work
01

Answers that cite their sources

Every answer shows where it came from. When the answer isn’t in your sources, the assistant says “I don’t know” instead of guessing. It is how hiy.ai answers.

Illustration: a policy document with the cited passages highlighted, next to an assistant answer that points to them, and an answer that says it does not know. Sample data.
02

Image and video pipelines

Many models behind one studio, with checks, labels and cost control at every stage. It is the pipeline behind Lovino.

Illustration: an image job moving through a pipeline: prompt check, model, output check, AI label and storage, plus a failed video job whose credits were returned. Sample data.
03

Agents with a person in the loop

Agents do the routine steps themselves, and stop for approval before anything that needs judgement.

Illustration: an agent adding a new supplier. It checks the paperwork, prepares the record and asks a person to approve before it changes the finance system. Sample data.

Who we build for

The teams we build for. And the work AI can take off their desk.

  • E-commerce and retail

    Illustration: a product page with new AI product photos waiting for review. Sample data.

    Product photos and copy, on brand.

    Image generation for product shots, descriptions in your brand’s tone, and support that knows your policies.

  • Media and creators

    Illustration: a video with automatic captions on a timeline. Sample data.

    Images, video and captions at scale.

    Generation across many models, with automatic captions.

  • Customer support teams

    Illustration: a support chat that answers with a source and offers to hand over to the team. Sample data.

    Answers from your help centre.

    Assistants that answer from your own docs and pass the hard questions to your team.

  • Trust and safety

    Illustration: a review queue with a result and a reason for each item. Sample data.

    Checks you can explain.

    AI-content detection, upload checks and review queues, with the reason behind each result.

  • Professional services

    Illustration: a first draft built from two documents, with sources shown. Sample data.

    Search and drafting over your documents.

    Find the right clause, note or file fast, and start from a draft instead of a blank page.

  • Internal tools

    Illustration: an internal workflow where an agent updates the CRM and waits for approval. Sample data.

    Agents for the busywork.

    Agents that move work between your tools: forms, records, updates and reminders.

Selected work

Built, shipped and running.

Three pieces of AI software we designed and built, and what each one shows we can do.

01 · Image and video generation

Lovino

AI image and video generation. Type a prompt, pick a model, get the result.

The problem

People want AI images and video without learning a different tool for every model.

What we built

  • One studio for image and video, with many hosted models to choose from.
  • Layered safety checks on prompts, uploaded images and generated images.
  • A “Made with AI” label written into each generated file’s metadata.
  • Credits and subscriptions, with automatic refunds when a generation fails.
  • Image and video generation
  • Content safety
  • AI labelling
  • Payments
  • Mobile
Illustration of the Lovino studio: a prompt, a choice of image models, the results, and a Made with AI label on each file. Sample prompt.

Lovino

Image and video generation · Web app, and a mobile app built with Expo

02 · AI-content detection

sighting.ai

AI-content detection that explains each result.

The problem

People need to know whether something was made by AI, and why a tool thinks so.

What we built

  • Detection for text, images, video and code, on self-hosted models.
  • Reports that show their work: flagged passages, the signal that decided and a confidence range.
  • Checks for content credentials (C2PA) and watermarks.
  • Built to delete uploaded images and video once the analysis is done.
  • Computer vision
  • Text classification
  • Self-hosted models
  • Explainable results
Illustration of a sighting.ai text report: flagged passages, the deciding signal and a confidence range. Sample text.

sighting.ai

AI-content detection · Web app

03 · AI twins and support agents

hiy.ai

An AI twin and support agent that answers your audience from your own content.

The problem

Experts and support teams answer the same questions again and again.

What we built

  • An AI twin that learns from websites, YouTube transcripts and documents.
  • Answers in the owner’s voice, with the sources behind each answer.
  • Labelled as AI on every surface, including embeds on other websites.
  • Works across several model providers: Anthropic, OpenAI, Google and DeepSeek.
  • Chat over your data
  • Citations
  • Embeddable agents
  • Multi-model
Illustration of a hiy.ai agent embedded on a website. It answers in the owner’s voice, shows its sources, is labelled as AI, and admits when a topic isn’t covered. Sample site and text.

hiy.ai

AI twins and support agents · Web app and embeddable widget

How we work

Your data and tools go in. Working software comes out.

Four steps, from the first conversation to launch.

  1. 01

    Discovery

    We learn the problem, the data you have and what a good result looks like.

  2. 02

    Design

    We choose the models and the architecture, and agree scope and cost with you.

  3. 03

    Prototype

    You get a working version early, tested on real inputs, not a slide deck.

  4. 04

    Production

    We launch it on the web or in the app stores.

Diagram: documents, databases, APIs and tools, and images and video go into the software Foldox builds. A human review step checks the results that matter. Apps, agents and pipelines come out.

The difference

AI projects, done with care.The same project, two ways.

Before

A typical AI project

Often starts from a model, then looks for a problem.

Is often judged on a demo with a few hand-picked examples.

Often ends at hand-over, with nobody running it after launch.

Often ties you to one model provider.

After

Working with Foldox

Starts from your problem, and picks the model last.

Is judged on a working version you can try yourself.

Keeps going after launch, with the same team.

Isn’t tied to any one model provider.

Start a project

Models and tools

The right model for each task. Across providers and open models.

  • Anthropic Claude
  • OpenAI GPT
  • Google Gemini
  • DeepSeek
  • Black Forest Labs Flux
  • Google Imagen
  • Google Veo
  • Kling
  • Seedance
  • Ideogram
  • Recraft
  • Stability AI
  • Runway
  • Luma
  • MiniMax Hailuo
  • Wan
  • Lightricks LTX
  • WhisperX
  • ModernBERT
  • Next.js
  • React Native
  • Expo
  • TypeScript
  • Python
  • Supabase
  • Postgres
  • Vercel
  • Vercel AI Gateway
  • Railway
  • Replicate
  • Cloudflare R2
  • ONNX
  • C2PA
  • Stripe
  • Resend
  • Sentry

Names are trademarks of their owners.

Working with Foldox

Start small, or go all the way.

Three ways to work together.

Start a project

  • Prototype

    A small working version first, so you see what AI can and can’t do for you before you commit to more.

    • A clear goal and a small scope
    • A prototype you can try yourself
    • A plan for what comes next

  • Build and launch

    We design and build the full product, ready for your users on the web, iOS or Android.

    • Design, build and testing
    • Web, iOS and Android
    • Safety and privacy built in

  • Run and improve

    We stay on after launch and keep it healthy as your needs and the models change.

    • Monitoring and fixes
    • Model updates when a better one fits
    • Changes as your needs grow

Questions

Common questions. Short, straight answers.

Contact

Still have a question?

Ask us about your own project, by form or by email.

Start a projectcontact@foldox.com
Can you work on a product we already have?

Yes. We can add AI features to it, or fix and extend the AI that’s already there.

Do we need our data ready before we talk?

No. Looking at the data you have is part of Discovery, the first step.

Will people know they’re talking to an AI?

Yes. Assistants we build are labelled as AI, and generated images and video carry an AI label in the file itself.

How do we start?

Tell us about the problem and the data you have, through the contact form or by email. Discovery starts from there.

Security and privacy

Careful with your data.

  1. Registered in Ireland

    Foldox Limited is an Irish company and follows the EU GDPR.

  2. Your data stays yours

    We use client data only for your project, never to train shared models.

  3. Agreements first

    We sign an NDA and a data processing agreement before we see your data.

Start a project

Have an AI idea, or a product to fix?

Tell us about it. We reply to every message, usually within two working days.

Start a projectcontact@foldox.com