Is your AI-generated app leaking secrets?

​Claude Code wrote the app. Now what?

​Code generation is no longer the hard part. What comes after it is:

​• Publishing: picking a host, configuring domain, and redeploying after every change. Fine for one app, a second job at fifteen..
• Consistency: five devs with five sets of installed skills produce five differently behaving apps..
• Secrets: generated apps often call APIs straight from the browser, with endpoints and env vars visible in DevTools. Anything the client can write to can be tampered with..

​In this live session, we’ll show how DronaHQ MCP handles what happens after generation.

​What we’ll do live

​• Prompt an app in Claude Code (or Cursor) and publish it from the agent, with no separate hosting, domain, or redeploy loop.
• Version the app and roll a change out.
• Show how skills built around your design system keep output consistent, whoever prompts it.
• Walk through RBAC aligned code generation across multiple apps and teams.

​Who it’s for

​• Developers building internal tools and automations with Claude Code, Cursor, or similar agents.
• Engineering leads and ops teams whose AI-built apps are starting to sprawl.
• Founders and business teams who want AI-built apps to be shippable, not just demo-able.

​What you’ll leave with

​• A repeatable prompt → publish → version workflow.
• A checklist for auditing your existing generated apps for client-side secrets.
• A setup that takes minutes to try.

​Bring your own app. We’ll take questions live, and we’ll look at a couple of audience examples if you want them reviewed.