AI Integration Services · Florence, Italy

AI Integration
& LLM App Development
in Florence

We have built mobile apps since 2010, and now we put AI inside them. LLM features with OpenAI and Claude, RAG over your own data, AI agents and assistants, and automations that run in production, not just in a demo.

Discuss your AI project

What We Build

Practical AI, shipped.

LLM Features in Apps

Language models where your users already are.

We connect OpenAI, Claude and custom models to your iOS, Android and web apps: smart search, summarisation, content generation, conversational interfaces and assistants that understand context. Server-side integration keeps API keys off the device and lets you monitor usage and costs.

OpenAI and Claude API integration
Chat and voice assistants inside apps
Structured outputs and tool calling
Usage monitoring, rate limiting and fallbacks

RAG & Knowledge Systems

Answers grounded in your own data.

Retrieval-augmented generation (RAG) systems that let an LLM answer from your documents, product catalogues, manuals or databases. We design the ingestion pipeline, the embeddings and vector database, and the retrieval logic that keeps answers relevant and sourced.

Document ingestion and chunking
Embeddings and vector databases
Hybrid search and reranking
Source citations and access control

AI Agents & Automation

Workflows that run themselves.

AI agents and automations built with n8n and LangChain that read emails, classify requests, update your CRM and answer customers on WhatsApp, with a human in the loop where it matters. It is the same kind of system behind MyPorter, our AI WhatsApp assistant for short-term rentals.

WhatsApp and chat assistants
n8n and LangChain workflows
CRM, email and back-office integrations
Human-in-the-loop review

On-device Machine Learning

Intelligence that works offline.

Machine learning that runs directly on the phone with Core ML and ML Kit: image and text recognition, classification and personalisation with no network round trip and no user data leaving the device.

Core ML on iOS
ML Kit on Android
Image, text and barcode recognition
Private, offline inference

How We Work

Prototype first, then production.

01 · Discover

We map your workflows and data and pick the use cases where AI pays off, and the ones where it should not be used.
02 · Prototype

A working prototype on your real data to validate quality, latency and cost before committing to a full build.
03 · Build

Production integration in your apps and backend, with evaluation, monitoring and guardrails.
04 · Improve

We measure answer quality and usage in production and refine prompts, retrieval and models over time.

Why APPSTYX

From demo to product.

AI demos are easy. AI that works inside a real product, for real users, is engineering. That is what we have done since 2010.

APPSTYX is a software house in Florence, Italy. We have shipped 100+ native iOS and Android apps, web platforms and cloud backends, and today we build AI into them: LLM features, RAG systems, agents and on-device models. Our co-founder Federico Frappi leads AI integration alongside iOS architecture and full stack work.

We also build and run our own AI product. MyPorter is an AI WhatsApp assistant for short-term rentals, developed by APPSTYX, which provides and manages the technology and WhatsApp Business messaging infrastructure behind it. Need the product around the AI? See mobile app development and cloud & backend.

15+ Years Building Software

100+ Apps Shipped

500M+ Users Reached

Our AI Stack

Tools of the trade.

Models

OpenAI API
Claude API
Custom models
Embedding models

Orchestration

LangChain
n8n
AI agents
Tool calling

Data & Retrieval

Vector databases
RAG pipelines
Node.js & TypeScript
AWS & GCP

On-device

Core ML
ML Kit
Swift & SwiftUI
Kotlin & Jetpack Compose

Questions

Straight answers.

How can AI be integrated into an existing app?

Usually through your backend: the app calls your API, which calls the language model with the right context and returns the result. This keeps API keys off the device, lets you switch models later and gives you logging and cost control. We integrate AI into existing iOS, Android and web apps as well as new ones.

Which AI model should we use, OpenAI or Claude?

It depends on the task, quality requirements, latency, cost and where your data may be processed. We test candidate models on your real use case before choosing, and we design the integration so the model can be swapped later without rewriting the app.

What is RAG and do we need it?

Retrieval-augmented generation (RAG) lets a language model answer using your own documents and data instead of only what it learned in training. You need it when answers must reflect your products, policies or knowledge base and stay up to date.

Is our data safe when we use LLMs? What about GDPR?

It can be, with the right design. We minimise and pseudonymise personal data before it reaches a model, choose providers and regions that fit your GDPR obligations, use API settings that exclude your data from model training where the provider offers them, and keep access controls on retrieved documents. For a full review, see our cybersecurity services.

Can AI run directly on the phone?

Yes. With Core ML on iOS and ML Kit on Android, tasks such as image recognition, text recognition and classification run on the device, work offline and keep user data on the phone. Larger language models usually run in the cloud, and we often combine the two.

Can you build an AI assistant on WhatsApp?

Yes. We build AI assistants that talk to customers on WhatsApp and connect to your systems. Our own product MyPorter is an AI WhatsApp assistant for short-term rentals, and we provide and manage the WhatsApp Business messaging infrastructure behind it.

Let's Build With AI

Tell us what to automate.

Email

[email protected]

Location

Via San Morese 9, 50019
Sesto Fiorentino (FI) Italy