- Hello, good evening, are you open on Saturday?
- Hello.
- Yes, we are open on Saturdays from 8:30 AM to 7:00 PM.
- How much does the service cost? I'm near Virgen de Luján, 8th ring.
- To provide you with a quote, could you please share your location?
- Okay, I'll send it.
- Give me a moment to arrive.
- Alright, I'll be waiting.
- I'll give you the price shortly...
- The price of the service is USD 53.
Your business operates on WhatsApp 24/7 — without the customer noticing the difference
24/7 support via WhatsApp with natural messages across multiple lines, a "typing…" indicator, and realistic response times. Respond, quote, schedule, and escalate to a human when needed.
What you'll stop suffering from.
-
Slow or out-of-hours responses — lost sales
-
Overloaded teams always answering the same questions
-
Information scattered across PDFs and spreadsheets that no one can find quickly
-
Slow manual quotes, with errors and no traceability
-
Lack of recording and analysis of conversations
-
Expensive SaaS chatbot platforms with limited logic
45 minutes of waiting.
The bot resolves it in just a few minutes.
Copied verbatim from a real chat in Santa Cruz. The customer takes 45 minutes — the bot quotes instantly, asks for location, waits, and closes the price.
The customer writes at any hour
It doesn't matter if it's 9:33 PM. The bot responds in seconds, with a "typing…" indicator and natural timings — just like a person online.
The quoting agent springs into action
The router classifies the intent and directs it to the specialized agent. They identify that they need the location to calculate the fare and ask for it.
Closes the price — on its own
With the received location, the bot queries the pricing engine and responds with the final value. No human intervention, in the same chat.
This exchange took 45 minutes with a human responding manually.
The bot resolves it in just a few minutes.
7 detailed views.
View of the WhatsApp chat with the bot handling a real case
Control dashboard with KPIs, live conversations, and agent activity
Technical diagram of the complete stack with n8n, FastAPI, ChromaDB, and services
View of the multi-agent system with the central router and specialists
Everything the product can do.
Multi-Agent Orchestration
- Smart router that classifies intent with LLM
- Domain-specific agents (quotations, inquiries, claims, scheduling)
- Persistent conversational memory per user
- Automatic handoff to a human for urgency or negative tone
- Structured outputs with JSON Schema
- Mixed models: GPT-5.1 for reasoning, GPT-4.1-mini for quick tasks
RAG Knowledge Base
- ChromaDB as a local vector database
- Ingestion of PDFs, manuals, price lists, FAQs, policies
- Semantic search with MMR (Maximum Marginal Relevance)
- Citing internal sources in every response
- Updatable base without redeploy
WhatsApp Integration
- Official WhatsApp Business API
- Support for text, image, location, audio, and documents
- Multi-number — one backend for multiple lines
- Automatic audio transcription with Whisper
- Scheduled sending and outbound campaigns
Control Dashboard
- Live KPIs of conversations and automatic resolution
- Real-time conversation view
- Audit of each message: agent, model, time, cost
- Metrics by agent
- Web management of the knowledge base
- Export of conversations and metrics
Continuous Improvement Audit
- Independent audit workflow
- Failure pattern detection
- Suggestion of new RAG documents or new agents
- Continuous improvement process without manual intervention
Mcp Integration
- Included n8n-MCP server
- Integration with Cursor, Claude Code, Claude Desktop
- Creation and modification of workflows using natural language
What it's built with.
Why this and not something else.
-
Humanize the conversation: respond in several short messages, with an indicator "typing…" and natural delays between responses — the customer won't notice the difference
-
Real multi-agent, not a single-prompt bot
-
Auditable RAG with its own basis — documents do not go to third parties
-
100% on-premise / private cloud — WhatsApp data does not leave the client
-
Proprietary code without vendor lock-in
-
Automatic auditing — the system improves itself by detecting its gaps
-
Integration with AI IDEs (MCP) — extension communicating with an AI
-
Modular architecture — agents plug in/unplug without breaking the rest
-
Adaptable to any industry — same core, just change prompts and RAG
-
No monthly fees per conversation or per user
Who is already using it?
High-volume WhatsApp service company
Home services
Commercial and support channel on WhatsApp integrated with the client's back office. The bot provides quotes by consulting the pricing engine, answers FAQs, schedules services, and escalates to a human when necessary.
Something like this for your business?
A 30-minute call to see if your situation is a good fit — no strings attached, no sales pitch.