Kodevengers
AI-Powered Solutions

AI agents that book, sell and answer for your business, 24/7

Not slide decks. Five working products with recorded demos: a WhatsApp receptionist that fills your calendar, an ordering agent for restaurant chains, a Shopify sales agent that sells in Instagram DMs, a property agent that books viewings, and a concierge that answers only from your own knowledge. Each one ships with a human handoff and a dashboard your team will use.

Watch the demos
AI Receptionist

24/7 WhatsApp booking agent

Watch demo
AI Ordering Agent

WhatsApp ordering for restaurant chains

Watch demo
AI Shopify Sales Agent

Selling and delivery on WhatsApp, Instagram and Messenger

Watch demo
AI Real Estate Agent

WhatsApp agent for property enquiries and viewings

Watch demo
AI Concierge

Grounded sales assistant with a human-in-the-loop knowledge base

Watch demo
Customer on WhatsApp
AI agent
Availability & booking engine
Calendar, reminders, inbox
WhatsAppDemo: A1 Luxury Nail & Spa, NY

AI Receptionist

24/7 WhatsApp booking agent

Never miss a booking again. It answers, qualifies, books, reminds, and hands off to a human when it should.

A WhatsApp receptionist for appointment-based businesses. Customers message the number they already use, and the agent answers questions from the business FAQ, books, reschedules and cancels appointments against live availability, sends reminders, and escalates to staff the moment someone asks for a person. The demo runs on the real service menu of A1 Luxury Nail & Spa in New York, whose website we also built. One engine, with dental, clinic, salon, physiotherapy, veterinary, chiropractic, and nail spa packs ready to switch on.

24/7

Replies, including weekends

7

Industry packs ready

1 chat

Book, reschedule, cancel

  • Books against real availability. Checks open slots, prevents double bookings at the database level, and confirms in chat.
  • Reminders that act. Sends 24h and 2h reminders. Customers reschedule or cancel by tapping a button in the reminder.
  • Guardrails built in. Refuses medical advice, admits when it does not know, and never invents a policy.
  • Human handoff. "Can I speak to a person?" pauses the AI, flags the chat in the inbox, and lets staff reply, then hand back.
  • Owner dashboard. Inbox, calendar, contacts, and analytics computed from real bookings, not estimates.
  • Live in minutes. An onboarding wizard takes a new business from sign-up to a working receptionist in one sitting.
DentalClinicsSalonsPhysiotherapyVeterinaryChiropracticNail spas

Stack: Claude · WhatsApp Cloud API · Next.js · Node · PostgreSQL · Redis queues · Google Calendar

Customer on WhatsApp
AI agent + 21 tools
Order engine
Branch, cart, pricing, promos
POS / kitchen display
WhatsApp

AI Ordering Agent

WhatsApp ordering for restaurant chains

Customers order in plain language. The AI talks, your order engine decides.

A WhatsApp-first ordering agent for restaurants and food chains. The model handles the conversation; a deterministic backend owns the menu, pricing, promotions, branch routing, delivery quotes and order status, so the AI can never misprice an item or apply an expired promo. It handles messy real orders ("two meals, one Coke one Sprite, make the Sprite large, no mayo"), upsells from rules, routes to the nearest open branch, and keeps the customer updated from kitchen to doorstep.

0

Pricing decisions made by the model

21

Tools wired to the order engine

4

Demo brands on one engine

  • Understands real orders. Per-item drinks and sizes, two corrections in one sentence, and Roman Urdu. Options resolve on the server, not in the prompt.
  • Backend owns the money. Cart, pricing, tax, delivery fee and promo eligibility are computed by code. The model only relays the numbers.
  • Branch routing. Typed address or shared location pin goes to the nearest open branch with the right fee and ETA.
  • Rule-based upsell. Offers one relevant add-on per conversation, with acceptance tracked for analytics.
  • Live order board. Staff advance orders from Preparing to Delivered. Every step lands as a WhatsApp message to the customer.
  • Reorder & takeover. "Same as last time" rebuilds a past order. Staff can take over any chat and hand it back.
Fast food chainsPizzaCloud kitchensCafésMulti-branch restaurants

Stack: Claude · WhatsApp Cloud API · Node · TypeScript · SQLite / PostgreSQL · Event bus for POS

Customer on WhatsApp or Instagram
AI agent + 16 tools
Shopify catalogue & orders
Courier booking & tracking
Staff dashboard
WhatsApp, Instagram, MessengerDemo: Sana Threads — a fictional demo store

AI Shopify Sales Agent

Selling and delivery on WhatsApp, Instagram and Messenger

Customers browse, order and track without leaving the chat. Every price, stock check and order comes straight from your Shopify store.

A sales and fulfilment agent for Shopify merchants that works inside WhatsApp, Instagram DMs and Messenger. It searches your live catalogue, builds the cart, collects delivery details, and creates a cash-on-delivery order in Shopify — but only after showing a full summary and getting an explicit yes. It then books the courier, writes the tracking number back to Shopify, and keeps the customer updated from booking to delivery. The model handles the conversation; the backend owns pricing, stock and every order decision. The demo runs on a fictional demo store so no real customer data appears.

3

Chat channels, one agent

16

Tools wired to Shopify & courier

0

Orders placed without confirmation

  • Sells from your real catalogue. Searches live Shopify products by category, colour, size and price, and checks stock per variant before promising anything.
  • No order without a yes. An order is only created after the full summary is shown and the customer explicitly confirms. The totals are re-checked at that moment, in code.
  • Cash-on-delivery into Shopify. Creates the order in your Shopify admin with the delivery charge and COD payment recorded, tagged by the channel it came from.
  • Courier booked automatically. Books the parcel, writes the tracking number back to Shopify, and messages the customer at every step from dispatch to delivery.
  • Cancellations by the rules. Cancellation is allowed before dispatch and refused after, and order details are never shown on Instagram without a matching phone number.
  • English, Urdu and Roman Urdu. Detects the customer's language and replies in it, including quick-reply buttons and delivery notifications.
Shopify storesApparel & lifestyleCash-on-delivery retailInstagram sellers

Stack: Claude · Shopify Admin GraphQL · WhatsApp Cloud API · NestJS · Next.js · PostgreSQL · Redis + BullMQ

Buyer on WhatsApp
AI agent + 12 tools
Matching & availability engine
Viewing booked + calendar
Agency dashboard
WhatsAppDemo: Lakeside Homes Realty — a demo agency in Austin, TX

AI Real Estate Agent

WhatsApp agent for property enquiries and viewings

Qualifies the buyer, matches them to your own listings, and books viewings against real availability — double bookings are impossible.

A WhatsApp agent for estate agencies. It captures what the buyer actually needs, searches only your verified listings, explains why each match fits, and offers viewing slots that are genuinely free — checked against agent hours, travel time between viewings and existing appointments. Booking requires an explicit confirmation, and the database itself blocks two viewings landing on the same agent or property. The demo shown here is the US configuration: dollars, square feet, beds and baths, in-person or virtual tours, and fair-housing-safe conversations. The same engine runs the Pakistani market in marla, kanal, lakh and crore, in English, Urdu or Roman Urdu.

0

Double bookings possible

12

Tools wired to listings & calendar

2

Markets: US and Pakistan

  • Qualifies before it books. Captures budget, area, size and timeline, and keeps track of what the customer actually said versus what the agent inferred.
  • Only your verified listings. It can only mention properties a search returned in that conversation, so it cannot invent a listing or a price.
  • Slots that are really free. Availability is computed from agency hours, agent calendars, blocked time and travel buffers between viewings.
  • Double bookings blocked. Two database constraints make it impossible for two viewings to take the same agent or property slot, even under simultaneous requests.
  • Fair-housing-safe in the US. Matching uses objective criteria only, and questions about neighbourhoods, mortgages or lease terms are handed to a person.
  • Handover with the full picture. Staff receive the requirements, the properties discussed, the open questions and the last message, then hand it back to the AI in one click.
Estate agenciesProperty developersRentals & lettingsUS and Pakistan markets

Stack: Claude · WhatsApp Cloud API · NestJS · Next.js · PostgreSQL · Redis + BullMQ · Google Calendar

Sitemap crawl + documents
Embeddings in pgvector
Retrieval + confidence gates
Cited answer or logged question
Partner answers feed back
Web chat, WhatsApp-readyBuilt for Tailyr

AI Concierge

Grounded sales assistant with a human-in-the-loop knowledge base

Answers only from your own stock, documents and team, with a citation on every claim.

Built for Tailyr, a multi-store AI concierge platform for high-value retail. Each store gets an assistant that answers strictly from that store's crawled website, uploaded documents and partner answers, cites every source, and refuses to guess. When it cannot answer, the question is logged for the store's team. Their answer is embedded back into the knowledge base instantly, so the assistant gets smarter with every question. Booking intent is detected up front and handed to a human as a lead.

4

Isolated stores on one platform

20/20

Adversarial attempts blocked

0

Hallucinations across the QA run

  • Strictly grounded. Two confidence gates block outside knowledge. Every answer carries numbered citations to the store's own pages.
  • Knowledge loop. Unanswered questions go to a partner dashboard. Answers are embedded back and attributed to the team member.
  • Store isolation. Multiple brands on one platform. A store can never see or leak another store's data.
  • Sold-stock intelligence. Knows what was previously sold for comparisons, but never offers it as available.
  • Manufacturer fallback. For models not in stock, quotes approved manufacturer pages and offers to source the item.
  • Adversarially tested. Sixty-question stress set covering prompt injection, pricing pressure and privacy probes, with zero hallucinations.
Luxury car dealersWatch retailersHigh-end audioYacht brokersAny multi-location retailer

Stack: RAG pipeline · pgvector · NestJS · Next.js · Supabase · OpenAI embeddings

How we build AI

Why these agents can run unattended

Most AI demos fall apart on the first real customer. Ours are designed around four rules that keep them safe when nobody is watching.

The AI talks. Your systems decide.

Prices, availability, policies and eligibility come from code and data you control. The model never gets to improvise on anything that costs money.

A human is always one message away.

Every agent detects "talk to a person", pauses itself, and hands the conversation to your team with full context. Staff hand it back when done.

Built where your customers already are.

WhatsApp first, web chat second. No app to install, no new login for your customers.

Measured on real outcomes.

Dashboards count bookings, orders, handoffs and unanswered questions from stored records, so you see what the agent actually did.

One team, whole product

AI agents work best inside a product people already use

An agent needs a booking page to send people to, an app to show order status in, a website that feeds its knowledge base. We build all three, so the AI is wired into your business instead of bolted onto the side.

  • Websites and e-commerce that feed the agent's knowledge and catch its leads
  • Mobile apps with the same booking, ordering and chat engine inside
  • The agent itself: prompts, tools, guardrails, dashboard and WhatsApp connection
Web portfolio Mobile app portfolio

From first call to live agent

  1. 01

    Use-case audit

    We map the conversations your team handles every day and pick the one with the highest volume and clearest rules.

  2. 02

    Working prototype

    A demo running on your real menu, services or knowledge base within two weeks, so you can judge it on your own data.

  3. 03

    Guardrails & integration

    Calendar, POS, CRM or website hooks, plus the refusal rules, handoff paths and logging that make it safe to run unattended.

  4. 04

    Go live & iterate

    WhatsApp number connected, staff trained on the dashboard, and weekly reviews of the conversations it could not handle.

Start with a working demo

See an agent running on your own data in two weeks

Tell us what your team answers, books or sells every day. We'll show you a prototype on your real menu, services or website before you commit to anything.

AI & Machine Learning Services