// For practitioners & consultants
These apps are easy to build.
We are happy to share how.
If you have expertise worth sharing — a methodology, a framework, a body of knowledge — you can turn it into a free, interactive AI app in an afternoon. Here is the process we follow.
1
Imagine the purpose
Start with one question: what should the user walk away with? A deeper understanding of themselves, a tailored framework, a decision made real through consequence? That answer determines everything else — the format, the flow, what the AI does.
2
Choose the interaction model
There are four patterns: a self-assessment with an AI-generated personal report; a discovery tool that applies a methodology to your situation; a scenario game where choices lead to consequences; or a meeting game — a quiz or a timer that energises a group. Pick the one that fits.
3
Assemble the know-how
This is what makes your app uniquely yours. Your methodology, your frameworks, your research. MEL Builder runs on BetterEvaluation's library. Tough Conversations draws on NVC, SCARF, and Radical Candor. The AI provides the interface — you provide the intelligence. Paste your knowledge into the conversation.
4
Ask Claude
Describe your objective, your format, and your methodology. Ask Claude to build it as a single-page HTML app — all the logic, design, and AI calls in one file. No server, no framework, no installation. Upload it anywhere and it is live — or simply open it in your browser desktop. Claude will handle the code; you focus on the content.
A note on running your app
While building and testing inside Claude, everything runs for free — no key needed. To share it as a standalone link, you will need an API key: ask Claude how to get one, it takes two minutes. Paste it into the file and you are live. The key will be visible in the code — fine for demos and pilots, just set a spending limit on your account. When you go public at scale, ask Claude to add a proxy server to keep it hidden.
5
Polish until perfect
Test it yourself. Note what feels wrong, what is missing, what the AI gets right and where it falls flat. Give specific feedback and keep asking Claude to revise. Design comes last — once the logic works, describe your look and feel. Iterate until the output genuinely surprises you. That is when you know it is ready.
Made something? We would love to feature it here.