+Human
Tell us about your project
AI development · custom software

We build custom AI.

Custom software for Italian companies and AI products of our own, with a modern stack and weekly sprints. From proof-of-concept to production, with the code yours from day one.

Tell us about your projectWhat we build
  • 2–4weeks to an MVP you can test with real users
  • 1sprint a week, with a demo at the end
  • Yoursthe code, with repo and documentation, from day one

  • Custom software

    Web platforms, internal apps, integrations, MVPs for startups. From brief to deploy, with your code and clean documentation.

    Avg. lead time6–12 wks

  • Our own products

    Seven products built with the same method we use for clients: the proof that we put AI into production.

    Products of our own7

  • A modern stack, iterative

    TypeScript, Python, Postgres, edge runtime, hosted & on-prem LLMs. Weekly sprints, continuous demos, code delivered with repo and CI.

    Sprints1/wk

What we build for you.

Six kinds of engagement, from a one-week sprint to a system in production. Every deliverable includes code, repo, documentation and training for your in-house team.

  • Custom software

    Web platforms & applications

    Lead time
    6–12 wks
    Stack
    TS / Python
    Output
    Repo + CI

    Internal web apps, client portals, dashboards, integrations with your systems (ERP, CRM, management software). Modern stack, your code, clean documentation.

    For whom Companies with a clear problem and processes that today live in Excel sheets or scattered tools.

    Ask for a brief

  • AI integration

    LLM, RAG & agent integration

    Lead time
    4–8 wks
    Models
    Hosted / on-prem
    Output
    API + eval suite

    Integrating LLMs into your workflows: semantic search over documents, agents that close tasks, conversational automation. With an eval suite and guardrails.

    For whom Companies with large document corpora or high-volume conversational processes.

    Ask for a brief

  • MVP & Proof-of-Concept

    An MVP in 4 weeks

    Lead time
    2–4 wks
    Output
    App + users
    Iter.
    3 sprints

    From an idea to a product real users can test in a month. For pre-seed startups, internal R&D, validating new lines of business.

    For whom Founders with a strong thesis who need to put the product in front of 10 real customers — fast.

    Ask for a brief

  • Intelligent automation

    Process automation

    Lead time
    2–6 wks
    Tools
    n8n / Make / API
    Output
    Workflow + docs

    We connect your tools (CRM, management software, email, drive, AI) into workflows that run on their own. Setup, monitoring, runbook for the in-house team.

    For whom SMEs with operational teams that today copy-paste data across 5 different apps.

    Ask for a brief

  • Code audit

    Code & architecture audit

    Lead time
    2–3 wks
    Type
    Assessment
    Output
    20–40pp report

    Independent audit of an existing system: architecture, code quality, security, scalability, technical debt. Report with priorities and an investment estimate.

    For whom Investors in tech due diligence, boards that inherited a legacy system, CTOs who want a second opinion.

    Ask for a brief

  • Maintenance & ops

    Maintenance & evolution

    Duration
    12 months
    SLA
    8h / month
    Output
    Quarterly roadmap

    Retainer for software already in production: bug fixes, small features, dependency updates, observability, best-effort on-call. So you don't stall.

    For whom Companies that built software 1–2 years ago and don't have a structured in-house tech team.

    Ask for a brief

We don't have a price list · every project is assessed after a free 30-minute introductory call, based on your needs.

Code you own

Seven products of our own.

Built with the same method and the same code we write for you. Client projects stay under NDA.

  1. BandoBuilder.itEU grants · Live
  2. MediTrace.aiMedical research, Mac · Early access
  3. QuietaAttention, iOS · Launching November 2026
  4. +Human ExecAI agent · Runs on your computer
  5. Agent BellAI agents, macOS · In daily use
  6. TrainingRights.itRights for AI training
  7. Dual Use ComplianceExport control, macOS

And two open-source projects: MiroFish-IT (AGPL-3.0) and AI-Audit Skills (Apache 2.0).

All products

Boring technology, modern integrations.

We don't chase the latest framework. We pick mature tools that stay in production for 5 years. The innovation is in the integrations and the AI models — everything else has to be predictable.

  • Frontend

    • TypeScript
    • React / Next.js
    • Vue / Nuxt (legacy)
    • Tailwind · design tokens
    • Astro for content sites

  • Backend & data

    • Python · FastAPI
    • Node.js · Hono / tRPC
    • Postgres · pgvector
    • Redis · queues
    • S3 / R2 storage

  • AI / LLM

    • Claude, GPT-4, Gemini
    • Open-source: Llama, Mistral
    • RAG · pgvector / Qdrant
    • Eval: LangSmith / Braintrust
    • Guardrails · PII redaction

  • Infra & ops

    • Vercel · Cloudflare
    • Fly.io · Railway
    • AWS · GCP (corporate)
    • GitHub Actions · CI / CD
    • Sentry · PostHog · logs

Every project is delivered with a GitHub repo you own and documentation.

On-prem hosting available for sensitive data

Brief, Design, Build, Ship.

Four phases, weekly sprints, continuous demos. No 80-page proposals, no black-box quotes: every week you see what we've built and decide whether to continue.

  1. PHASE 01

    Brief

    A free 30-minute scoping call + a one-page written brief: problem, constraints, expected output, definition of "done". If the fit isn't there, we tell you.

    1 page · fixed scope
  2. PHASE 02

    Design

    Technical architecture, stack choices, interface mockups, sprint breakdown. Decisions that stay written down — no "we said it out loud".

    Architecture · mockups
  3. PHASE 03

    Build

    Weekly sprints with demos. Code written, tested, in continuous integration. You have repo access from day 1: no surprises at delivery.

    Sprints · weekly demos
  4. PHASE 04

    Ship

    Production deploy, in-house team training, operational runbook, 30 days of best-effort on-call. The code is yours, the repo is yours, the knowledge too.

    Deploy · handover · docs

The questions we always get.

Eight honest answers about the way we work: costs, timelines, code ownership, maintenance. If you can't find yours, write to us — we reply within 24h.

How much does a custom project cost?

We don't have a price list: every project is assessed after an introductory call, based on your needs. For a sense of timing: MVP 2-4 weeks, custom platform 6-12 weeks, LLM/RAG integration 4-8 weeks.

The introductory call is our free 30-minute scoping call: you tell us the problem and, if we can help, we send you a one-page fixed-scope brief with what we do, when and how much. No 80-page proposals.

How long from brief to delivery?

MVP / PoC: 2-4 weeks. Custom platform: 6-12 weeks. AI integration: 4-8 weeks. We work in weekly sprints with demos, so halfway through you already have a working version in hand — you don't wait until the end to see something.

Who owns the code?

You do. Always. A GitHub repo you own from day 1, read access for the whole team. Technical documentation, operational runbook, knowledge transfer to the in-house team at delivery. If tomorrow you decide to bring the project in-house, we're out in two days.

What stack do you use?

TypeScript + Python by default. React/Next.js on the frontend, Node.js or FastAPI on the backend, Postgres as the database. For AI: Claude/GPT/Gemini in the cloud, Llama/Mistral self-hosted where on-prem data is required. All on Vercel/Cloudflare/Fly or your own cloud (AWS/GCP/Azure) if you prefer.

We don't chase the latest framework: we pick tools that stay in production for 5 years and that anyone can maintain.

And after delivery? Do you maintain it?

Three options. (1) Handover: we hand everything to your in-house team and stay available for 30 days for questions. (2) Monthly retainer: you keep an open channel with us for bugs, small features, updates; the fee is set after an introductory call, based on your needs. (3) Mixed: your team handles the day-to-day, we're on-call for the big things.

Do you work for equity or only for a fee?

Only for a fee, save very rare cases. We believe the equity-for-build model muddles the incentive: you need us to ship the product, not to push metrics for the next valuation. For pre-seed startups we offer a fixed-scope MVP, assessed after an introductory call based on your needs — so you test the product before diluting.

Can I see projects you’ve shipped?

Our products are public: BandoBuilder.it live, MediTrace.ai in early access, Quieta launching on iOS, +Human Exec, Agent Bell, TrainingRights.it and Dual Use Compliance; MiroFish-IT is open source on GitHub. Client projects are under NDA and we don’t show them without authorization.

Is our users' data safe?

Yes, and it's one of the reasons we exist. For AI projects: PII redaction, configurable log retention, on-prem option (Llama/Mistral self-hosted) for sensitive data. For GDPR and the AI Act we're specialists — we also handle compliance audits separately, so we know where to look.

Got an idea? Let’s build it together.

A free 30-minute scoping call: you tell us the problem, we tell you whether we can solve it and — if so — how much time and budget it takes. A one-page brief within 48h of the call.

Studio
Via Francesco Cherubini 6, 20145 Milan