Live Agent System
The hardest job for a new AI studio is earning trust: how do you prove capability before you have cases? EHNUH solved it with the site itself — the live agent you use while reading this is not a mockup, but a real model stack running at zero monthly cost.Try the live demo on the homepage
YEAR
2026
TAGS
Duration
Live — continuously improving
Team
EHNUH AI
Stack
PROBLEM
The paradox of a new AI studio: you need proof to win work, and work to build proof. Fake cases and inflated metrics collapse on the first real client. Plus a technical constraint: zero budget, no payment card, instant replies, resilient failover, abuse protection.
APPROACH
We stopped talking and started showing. On the homepage, visitors describe a task and watch the agent work — honestly labeled steps, real model responses. Multi-layer failover: six stages across two independent gateways and three separate model vendors. Quota runs out at the account level, not per model, so the chain switches gateways as well as models. Anonymous conversation logging, crash alerts, bot protection.
RESULT
The case study is the system running as you read: not a mockup, a live model. · Monthly cost: ₺0 on free tiers. · Six-stage failover: two independent gateways, three model vendors. · Anonymous demand logging. · Bot-protected quota. · Automatic failure alerts. Lesson: well-built AI speaks through a running, measurable system.