Your business answering 24/7Message me on WhatsApp →

Case study · Hospitality · Valencia

A tapas bar that takes bookings on its own while the team works the bar.

La Esquinita del Turia is Dadinirt Crispin's Venezuelan tapas and sandwich bar next to the Roig Arena. I built its website, an AI agent that handles bookings, and the tools that keep the staff informed from their phones.

Project

Client
La Esquinita del Turia
Owner
Dadinirt Crispin
Sector
Hospitality · tapas bar
Where
Valencia, Spain
Live
Website: November 2025 · Agent: September 2026

The challenge

Bookings came in by phone, right at rush hour.


In a small bar, whoever answers the phone is also serving. Every booking call is a dish left waiting, and every rushed note is a table that might get double-booked.

The menu changes too, and weekend empanadas are morning-only: customers kept asking the same questions by phone and WhatsApp.

What the bar needed

  • Take bookings without picking up the phone
  • Never give the same table twice
  • Know about every booking instantly
  • See what people look at on the menu

What I built

An agent, a dashboard and phone alerts.


AI agent

Booking agent on the website

Customers pick day, time and party size with buttons and a calendar; a chat handles questions. The agent checks real availability before confirming and, if full, offers the nearest free slot.

No double bookings

The database decides, not the AI

The agent talks, but a database function assigns the table by zone and capacity. The AI can never hand out a table that is already taken.

Alerts

Push notifications to staff

Every new booking arrives as a push notification on the team's phones and by email. Same-day bookings close in time for the kitchen to plan.

Dashboard

Booking dashboard with staff accounts

Each team member logs in with their own account. New bookings stand out and past ones collapse, so the day reads at a glance.

Protection

Abuse limits

The chat is public, so it defends itself: booking limits per phone and IP, short conversations scoped to booking. Nobody can flood the bar with fake bookings.

Metrics

Menu analytics without cookies

An in-house counter records which dishes people look at, with no cookies or personal data. Every morning a five-line email sums up yesterday's visits and top dishes.

Also

A website built to be found.


Pages for local searches

Dedicated pages for people looking for a bar near the Roig Arena, tapas in Quatre Carreres or a place to eat next to El Saler shopping centre.

Menu with allergens

Each dish shows its allergens as colour icons with a legend, reviewed with the kitchen before going live.

Where it stands

Live, and measured from day one.


The website is live and the booking agent launched in September 2026. Everything that happens — bookings, menu views, contacts — is recorded from day one, so the bar knows what works instead of guessing.

The agent is ready to serve more than one venue: the next bar can be onboarded without touching the first one's website.

Stack

  • Agent built on Claude (Anthropic) with custom tools
  • Postgres database (Supabase)
  • Push notifications and email
  • Next.js website on Vercel

Run a bar or restaurant?

The same booking agent can work for you.

It adapts to your tables, hours and way of working, and drops into your current website with a single line.

See the booking agent

Which process is eating your team's time?

I'll show you a demo with your case and tell you what can be automated, what can't, and where to start.

Case: AI bookings at a tapas bar in Valencia | AppsConCerebro