AALAndrés Listorti
ES

Taking new projects: websites, apps & AI automations

AI-native web & app builder and lawyer

I build websites, custom business apps and AI automations with the latest AI tools, so you get a working product in days or weeks, not months.

Email Resume
Buenos Aires · GMT-3
orchestrator.run✓ 0↺ 0
  • claude-code: split the redesign into parallel tasks — approved, no two agents touch the same file
  • muse-spark: build the hero orchestration console — approved, pauses when off-screen
  • codex: mark order paid from webhook — returned, a missing header must fail closed
  • codex: fix: always verify the signature — approved, fails closed now
  • nemotron: canvas neural field for the hero — approved, static frame for reduced motion
  • claude-code: reserve stock during checkout — approved, no overselling under concurrency
years practicing business law
0+
years building software
0+
products running in production
0
AI coding agents in my daily loop
0

Claude Code, Next.js, React, Node.js, TypeScript, AI agents, Python, Firestore, Mercado Pago, MCP, OpenCode, Vercel, Nano Banana, Business law

Process

AI writes most of the first draft. My job is to make the task impossible to misread, route it to the right model, and refuse anything I wouldn't put my name on.

A developer at night directing floating AI agent terminals connected to one central node
  1. 01

    Brief

    Turn a request into a spec an agent can't misread: goal, constraints, files it may touch and what done looks like.

    • › SPEC.md + CLAUDE.md per project
    • › out-of-scope written down
    • › acceptance checks named up front
  2. 02

    Plan

    Choose the architecture and split the work into pieces that can run in parallel without colliding.

    • › one owner per file
    • › interfaces and types first
    • › pick the model per task
  3. 03

    Delegate

    Claude Code orchestrates. Implementation goes to the model that fits: Claude, Codex, or open models like Muse Spark and Nemotron through OpenCode.

    • › parallel agent sessions
    • › MCP tools: Vercel, Drive, image generation
    • › cheap models for boilerplate
  4. 04

    Review

    Nothing ships unread. Agent output gets reviewed like a pull request: correctness, edge cases, security and readability.

    • › tsc + build + tests must pass
    • › auth, secrets, webhooks checked by hand
    • › returned with notes when it falls short
  5. 05

    Ship

    Preview deploy per branch, checked in a real browser, then merged.

    • › Vercel preview on every push
    • › mobile + reduced-motion pass
    • › merge to main

AI work

What I've shipped lately with AI agents in the loop, and how the work was split between them and me.

  • Multi-agent build, 2026

    This portfolio

    I wrote the design spec and split it by file. Claude Code orchestrated; the hero console and scroll sequence went to Muse Spark, the canvas field and code viewer to Nemotron, both through OpenCode over MCP. I reviewed, type-checked and built every file before merging.

    Claude CodeOpenCodeMCPNext.jsMotion
    Source

  • AI agent SaaS, 2026

    Digital Assistant

    An agent that drafts Instagram posts and images with Gemini, OpenAI or Claude, schedules them and handles the inbox, but never publishes without a human approval.

    LLM APIsAgentsNode.jsMeta Graph API
    Product site

  • Daily setup

    Agent workstation

    Claude Code as orchestrator, with MCP connectors for Vercel, Google Drive, image generation and a local ComfyUI, plus OpenCode to route routine tasks to open models and keep costs down.

    Claude CodeMCPOpenCodeComfyUI

Services

For businesses in any niche, remote from Buenos Aires, in English or Spanish. Fixed price and clear delivery dates.

Websites & landing pages

Fast, mobile-first sites that load quickly and rank: from a one-page landing to a full company site on your own domain.

  • › live in days, not months
  • › SEO and analytics ready
  • › WhatsApp and contact forms wired in

Custom business apps

Apps shaped around how your business actually works: online stores, booking, pricing and stock tools, client portals and dashboards.

  • › login, database and admin panel
  • › payments with Mercado Pago or Stripe
  • › built for your niche

AI assistants & automations

Chatbots and agents that answer clients, capture leads and draft content, with a human approval step where it matters.

  • › website, Instagram or WhatsApp
  • › Gemini, OpenAI or Claude
  • › human-in-the-loop by design

AI visuals & legal-aware delivery

Product photos, social media assets and short promo videos made with AI, plus plain-language notes on data and terms of service from a practicing lawyer.

  • › Nano Banana and Higgsfield visuals
  • › privacy-aware by default (GDPR / Ley 25.326)
  • › bilingual EN / ES
Start a project

Projects

Zapatería Genaro — screenshot of the live site

Full-stack e-commerce, 2026

Zapatería Genaro

Online store for an Argentine handmade footwear brand: catalog, cart, customer accounts, admin panel and in-site payments.

Next.jsReact 19ExpressTypeScriptFirestoreMercado PagoVitest
Live Code
Problem
The brand needed its own store instead of relying on marketplaces, with payments and stock that stay consistent under real traffic.
Approach
A Next.js front end and a separate Express / TypeScript API, both on Vercel, with Firestore and Firebase Auth. Mercado Pago Checkout Bricks keeps payment inside the site.
Key decisions
  • Signature-verified, idempotent webhooks: an order is fulfilled exactly once, whichever path confirms the payment first.
  • Stock is reserved in a transaction for 30 minutes, and the payment link expires at the same moment.
  • Underpaid payments are never fulfilled; PayPal was switched off because its sandbox mode allowed fake payments.
  • Vitest + Testing Library, Sentry monitoring and transactional email via Resend.
Outcome
Live with catalog, accounts, admin and Mercado Pago checkout working end to end: payment inside the site, stock reservation and automatic order confirmation.
Digital Assistant — screenshot of the live site

AI agent SaaS for Instagram, 2026

Digital Assistant

Multi-tenant web app that turns an idea into an approved, published Instagram post, plus metrics and inbox handling.

AI agentsLLM APIsReactNode.jsExpressFirestoreMeta Graph API
Product site Private code · demo on request
Problem
Small businesses want AI-generated content without losing control of what goes out under their name.
Approach
A React front end and a Node / Express API on Firestore. Each post is one document that moves through a state machine: queued → generated → approved or scheduled → publishing → published.
Key decisions
  • Pluggable LLM providers for copy and images (Gemini, OpenAI, Claude), plus a manual mode.
  • Nothing is published through the Meta Graph API without an explicit human approval.
  • Errors send a post back to review instead of losing it or retrying blindly.
  • Workspaces, Google sign-in and monthly plans billed through Mercado Pago.
Outcome
In production at app.digitalassistant.com.ar, with review, calendar, comment and DM inboxes, and metrics.

More projects

Internal business tool, 2026

Costs & Pricing App

Installable PWA that replaced a footwear factory's pricing spreadsheet: per-article cost breakdowns, price levels, charts and Excel / PDF reports.

PWAJavaScriptFirebaseFirestore
Private code · demo on request

LLM chat interface, 2025

AI Chatbot

Streaming chat UI built with the Vercel AI SDK on top of OpenAI models, Next.js and TypeScript.

Vercel AI SDKOpenAINext.jsTypeScript
Code

Java back end, 2024

Spring Boot REST API

Layered e-commerce API (model / repository / service / controller) with full CRUD for users, customers, products and addresses.

JavaSpring BootJPAH2
Code

Node.js back end, 2025

Real-time Products API

Express API for products and carts with a live product view pushed over Socket.io.

Node.jsExpressSocket.ioHandlebars
Code

AI + Law

Most AI risk isn't in the model; it's in what the product does with people's data and what it's allowed to do on their behalf. That's where my two careers overlap.

Scales of justice drawn as a glowing knowledge graph, weighing legal papers against code

Human in the loop, by design

In Digital Assistant, nothing an LLM writes is published without an explicit human approval, and failed posts go back to review instead of being retried blindly. Accountability is a product decision, not a disclaimer.

Data-protection-aware development

I design with Argentina's Personal Data Protection Law (Ley 25.326) and GDPR basics in mind: collect only what's needed, know why each field exists, keep secrets and credentials out of repos, and make deletion possible.

Contract and ToS literacy

I read API terms, platform policies and client contracts the way I read code: what's promised, what's excluded, who carries the risk. That matters when a product depends on Meta, payment providers or model vendors.

Governance you can explain

Clients and compliance teams need to understand what an AI feature does. I document behavior, limits and failure modes in plain language, in English and Spanish.

I'm a practicing lawyer in Argentina, not a certified privacy officer; for regulated work I collaborate with the client's legal team.

About

Portrait of Andrés Listorti

Andrés Antonio Listorti

Buenos Aires · GMT-3

I've practiced business law since 2010 and have been building digital products since 2020. Today I work AI-native: AI tools do most of the building, and I own the result, from the plan to the product running live. I build for businesses in any niche: stores, services, professionals and startups.

The legal background isn't a footnote. It's why I scope work carefully, think about what can go wrong, and explain decisions clearly to clients who aren't technical.

Law
Since 2010
Business law · Universidad de Belgrano
Building
Since 2020
AI-native · React & Next.js certs
Languages
English C1 · Spanish native
Works with
Next.js · Node · Java/Spring · Firebase

Two careers, one way of working.

  1. 2020 — present

    AI-Native Web & App Builder (freelance)

    Self-employed · Remote

    Digital products for small businesses, built AI-native: the Zapatería Genaro online store, a costs & pricing tool for a shoe factory, and Digital Assistant, an AI SaaS for Instagram.

  2. 2010 — present

    Independent Attorney

    Business, commercial, labor & civil law

    Independent practice including commercial law for banking clients. Where I learned rigorous reading, risk analysis and clear communication with non-technical clients.

Education

  • Full Stack Development

    Coderhouse, 2021 — 2024

  • Next.js — Intensive Career (cert.)

    Coderhouse, 2025

  • React JS — Intensive Career (cert.)

    Coderhouse, 2024

  • Algorithms in Python

    ITMaster Academy, 2022 — 2023

  • Lawyer — Business Law

    Universidad de Belgrano, 2004 — 2009

Tools I've actually shipped with.

Front end
TypeScriptJavaScriptReactNext.jsTailwind CSSHTML / CSS
Back end
Node.jsExpressJavaSpring BootJPAPythonSocket.io
Data & services
FirestoreFirebase AuthMySQLMercado PagoMeta Graph APIResendSentry
AI
Claude CodeGitHub CopilotOpenAI CodexCursorVercel AI SDKOpenAI / Gemini / Claude APIs
Tooling
Git & GitHubVercelVitestPostmanFigmaJira / Scrum

Contact

Pick whichever channel suits you. I answer in English or Spanish, usually within a day.

Download resume (PDF)