19 active · 7 in development

26 AI agents.
Working 24/7.

While you sleep, your agents analyze sales, detect fraud, forecast the close, optimize your menu and report everything to you on Telegram. No other POS has this.

554+Autonomous actions
98.2%Uptime
24/7Continuous monitoring
$0Extra cost

Architecture

An autonomous nervous system

The agents aren't isolated features. They're a connected ecosystem that talks to itself, learns from your data and acts without you having to do a thing.

Data sources
POS (Wansoft)
Reservations
Google Reviews
WhatsApp
Core engine
Orchestrator
Supabase (data)
Groq (reasoning)
Tentacles (26 agents)
Reports (4)
Operations (5)
Optimization (5)
KB 24/7 (2)
Suppliers (3)
Coming soon (7)
Output
Telegram
Dashboard
WhatsApp

The 26 agents

Every agent, a specialist

Organized into 6 categories. Each one with a defined trigger, schedule and output. Zero configuration on your end.

Reports -- 4 agents
Active daily7:00 AM
Daily briefing
Executive summary at 7AM: yesterday's sales, today's reservations, comparison vs last week, 3 priority actions.
You closed yesterday at $18,500. +12% vs last Tuesday. You have 2 garden events today. Action: order more corn bread, it sold out at 2PM.
WeeklyMon 9:00 AM
Weekly report
Full analysis of the week: sales trends, top servers, categories that grew or fell, average ticket.
Week 21: $128,000 (+8.3% vs W20). Chilaquiles still #1. Bowls up 23%. Carlos sold $31K. Average tips: 9.2%.
Active intradayevery 2h
Intraday sales
Sales updates throughout the day. Hour-by-hour comparison vs the same day last week.
2:00 PM update: $9,210 sold. You're 15% ahead of last Thursday. 23 tickets, $400 average ticket. Peak hour: 1PM.
Active intraday2PM / 4PM / 6PM
Close predictor
Predicts the day's close using historical data and your current sales pace. Updates 3 times per shift.
4PM forecast: you'll close at ~$21,500 (87% confidence). If you match last Thursday's 4-7PM pace, you could hit $23K.
Operations -- 5 agents
Active intraday2PM / 4PM / 6PM
Anomaly detector
Detects metrics outside the historical pattern: unusually low sales, discount spikes, categories dropping without reason.
ALERT: Discounts today $2,100 (3.2x the average). 4 of 6 came from the same server. Review Andres' comps.
WeeklyFriday 8AM
Anti-fraud
Analyzes comp, cancellation and discount patterns. Flags servers with out-of-range ratios and cross-checks against tips.
Andres has 14 cancellations this week (team average: 3). 8 were "wrong dish". Comp-to-sales ratio: 6.2% (team: 1.8%).
Active daily7AM / 4PM / 7PM
Kitchen quality
Monitors kitchen cancellations and returns vs baseline. Detects when a dish is failing systematically.
Enchiladas suizas: 3 returns today (baseline: 0.2/day). Possible ingredient issue. Check the salsa verde.
Active daily7AM / 4PM / 7PM
Table time
Analyzes table turnover: average occupancy time, slow tables, bottlenecks during peak hours.
Average turnover: 47 min. Table 8 sat for 92 min (outlier). At peak you lost ~3 tickets to occupied tables not ordering.
Active daily7AM
Config validator
Validates that your system configuration is correct: prices, schedules, integrations, data sync, permissions.
OK: All configs valid. Wansoft sync: 12 min ago. 0 permission alerts. Menu up to date.
Optimization -- 5 agents
WeeklyMonday 9AM
Menu engineering
Classifies every dish on the BCG matrix: stars (high sales, high margin), cows (high sales, low margin), puzzles and dogs.
Stars: Chilaquiles ($18K/wk, 68% margin), Avocado Toast ($9K, 72%). Dogs: Ceviche ($800, 31%). Action: raise the acai bowl price $15.
WeeklyMonday 9AM
Staffing optimizer
Analyzes demand patterns by day and hour. Suggests optimal schedules for each server based on real data.
On Mondays you need 3 servers (not 5). Saturdays you need 6 from 11AM-3PM. Carlos and Laura perform best on the AM shift.
Active intraday2PM / 4PM / 6PM
Upselling agent
Spots upselling opportunities per server: who sells the most drinks, extras, desserts. Suggests personalized tactics.
Laura only sold 2 desserts today (team average: 5). If she offers coffee with dessert like Carlos, she could add $600/day.
WeeklyFriday 8AM
Tips analyzer
Tip analysis by server and over time. Correlates tips with service type, shift and section served.
Carlos: 12.1% average (best on the team). Andres: 6.8% (lowest). Tips run 34% higher on the terrace vs the dining room.
WeeklyMonday 9AM
Waste detector
Detects waste and shrinkage by analyzing gaps between theoretical and actual inventory. Alerts when an ingredient is consumed more than expected.
Avocado: actual usage 23kg, theoretical 18kg. 27% shrinkage. Likely cause: inconsistent portions on toasts and bowls.
Knowledge 24/7 -- 2 agents
Active 24/7on-demand
Knowledge Base (KB)
Natural-language questions over Telegram. Answers with real POS data in seconds. No dashboard to open.
You: "how much did we sell yesterday?" -- KB: "Yesterday $18,400. 42 tickets. Average ticket $438. Top: Chilaquiles ($3,200). Carlos sold $5,100."
Active 24/7on-demand
Orchestrator
The central router. Receives every Telegram message, classifies the intent with AI and dispatches it to the right agent automatically.
Message received: "how are we doing today?" -- Classified: kb/sales -- Dispatched to: wansoft-query -- Answered in 4.2s.
Suppliers and integrations -- 3 agents
WeeklyMonday 9AM
Suppliers agent
Analyzes supplier costs, detects price hikes and suggests renegotiations based on your historical purchase volume.
Oaxaca cheese went up 18% this month. With your volume (12kg/wk) you can negotiate a fixed quarterly price. Contact the supplier.
Active daily8:00 AM
Wansoft staleness
Verifies that POS data sync is up to date. Alerts if the data is more than 24 hours behind.
Sync OK: last data point 43 minutes ago. All up to date.
Active daily10:00 AM
Pending reservations
Detects unconfirmed reservations or ones missing a contact phone. Alerts you to follow up before it's too late.
2 pending reservations for this weekend: AMA-5102 (Saturday, unconfirmed) and AMA-5105 (Sunday, no phone).
Coming soon -- 7 agents in development
In developmentQ3 2026
Reviews monitor
Monitors Google Maps reviews in real time. Alerts on negative reviews and drafts AI-generated responses.
New 3-star review: "Food was fine but it took forever." Response draft ready for your approval.
In developmentQ3 2026
WhatsApp bot
Conversational bot for reservations, menu questions and automatic confirmations. Integrated with the CRM.
Customer: "I want to book for 8 people on Saturday" -- Bot: "The Garden is available at 2PM. Shall I confirm?"
In developmentQ3 2026
CRM agent
Identifies frequent customers, calculates lifetime value and suggests personalized retention actions.
Maria Lopez hasn't visited in 3 weeks (she used to come every 5 days). Send a "we miss you" message with a 10% discount.
In developmentQ4 2026
Weather agent
Cross-references the weather forecast with historical sales patterns. Adjusts forecasts and suggests inventory prep.
Rain all day tomorrow. Historically you sell 22% less in the rain. Suggest cutting salad prep and boosting soups.
In developmentQ4 2026
Hermes (delivery)
Optimizes delivery operations: prep times, routes, performance by platform (Uber, Rappi, Didi).
Uber Eats: $320 average ticket, 18 min prep time. Rappi: $280, 24 min. Suggest prioritizing Uber at peak hours.
In developmentQ4 2026
Kitchen agent
Monitors kitchen prep times via the KDS. Detects bottlenecks and optimizes order flow.
Average time: 14 min. Cold station: 8 min (OK). Grill: 22 min (bottleneck). 4 orders in queue.
In developmentQ4 2026
Marketing agent
Generates social media content from your data: star dishes, events, smart offers.
Your chilaquiles sold $18K this week. Suggested post: "Monterrey's #1 dish. 127 orders this week."

How it works

From installation to agents with a clear path

1

We connect your POS

We integrate your current data. No hardware changes. Not a single minute of downtime for your operation.

2

We activate the agents

We configure the 19 active agents with your schedules, team, menu and goals. Fully personalized.

3

You connect Telegram

A bot in your Telegram. Ask anything in natural language. Reports arrive on their own.

4

You run on autopilot

The agents work 24/7. You just read the alerts and make decisions with data, not gut feeling.

Ready to start

26 agents working for your restaurant

Book a 15-minute call. We show you the agents running live with real data from a restaurant in Monterrey.