What it is
A chat simulator to train agents to handle multiple chats. Run structured scenarios to build speed and accuracy before agents hit the floor.
What kept happening
Chatbots getting deployed without anyone systematically testing the edges. The gaps show up in production — in escalations, in complaints, in the tickets that shouldn't exist. By then it's someone else's problem to explain.
What I built
Structured prompt sets by category, run against any chat endpoint, scored and logged per session. The audit you run before launch — or after, when something breaks and you need to know why.
Stack
Next.js 14TailwindClaude APIVercel
What I'd change
Scoring is currently manual interpretation — no automated pass/fail rubric yet. That's the next build.