Inovais
Customer service and sales agent on WhatsApp and Telegram, adaptable to any industry.
TypeScript / LangGraph / RAG / Multi-LLM / WhatsApp API
The problem
The customer asks at night, the business replies the next day and the sale is gone. During business hours, the team repeats the same answers. And conversational service shouldn't be rebuilt from scratch for every client.
The solution
Inovais answers on WhatsApp and Telegram through a 7-node LangGraph flow: it understands intent, searches the catalog with RAG, recommends and walks through the negotiation in conversation. Multi-LLM routing with circuit breaker, prompt-injection detection and LGPD across the whole flow. You swap the context — dealership, restaurant, clinic, real estate — not the engineering.
The proof
Over 1,100 automated tests hold the agent's behavior in place. The demo runs on a public sandbox as the agent of a used-car dealer — try it directly on Telegram, nothing to install. Switch industries in the simulation and watch the agent change context.
NARRATED DEMO
Step-by-step walkthrough of the system in use — recorded on the live environment.
SEE IT WORKING
Switch the industry and watch the same agent change context. The conversations follow short demo scripts.
Do you have vegetarian options?
In the simulation: we have 3 vegetarian dishes on today's menu. Want me to send photos and prices?
Book a table for 4 tonight at 8pm
In the simulation: table for 4 at 8pm booked under the name you give. Confirmation sent right here.
Same agent, different context: industry, tone and answers change; the service structure stays the same. Fictional scripts.
Want something like this for your business?
In three steps the house reviews your case and issues the report — no strings attached.