Coding Agent Cloud How it works Pricing For agencies About us Contact Careers Blog Login Start your coding agent

Guide

From Vibe Coding to Production: What No One Tells You About Shipping AI-Built Apps

The five gaps between "it works in the builder" and "real users can use it". How Lovable, Bolt, Cursor and Claude Code each handle each gap, and why human in the loop is the layer no AI tool ships with.

The five gaps no one warns you about

Gap 1: Security

The builder has your test API keys in the bundle. Production needs them in env vars. Your Supabase has RLS off. Production needs it on. Your auth endpoint has no rate limit. Production needs one. Read more in our vibe coding security guide.

Gap 2: Environment isolation

The builder uses one database for everything. Production needs separate development, staging, and production databases. Otherwise a buggy feature wipes real customer data.

Gap 3: CI/CD

Builder has "publish". Production needs version control, branches, code review, automated tests where possible, deploy gates. Otherwise rolling back means rebuilding from memory.

Gap 4: Monitoring

Builder shows the preview as you click. Production runs unattended. You need error tracking (Sentry), uptime monitoring (BetterUptime), and alerts that wake you when things break. Otherwise users tell you about issues you should have known first.

Gap 5: Scaling

Builder hosts five concurrent users fine. Production might see five hundred at launch. Builder native hosting often does not scale. Self serve PaaS scales but at exponential cost. Managed deployment scales predictably with one bill.

How each AI tool scores on each gap

GapLovableBoltCursorClaude Code
Security defaultsSome warningsNoneUp to userUp to user
Environment isolationLovable Cloud onlyNone defaultManualManual
CI/CDGitHub sync, basicManual setupManualManual
MonitoringNone defaultNone defaultNoneNone
Scaling storyLovable Cloud limitsNetlify limitsNoneNone

Why human in the loop closes all five

An engineer reviewing each release does what no automated tool does: reasons about the gaps holistically. They notice that you forgot the production env vars, that monitoring is not configured, that the database has no backups. Read more about human in the loop AI coding.

For the operational version of closing these gaps, see managed deployment for AI apps. For the DIY version, follow our production readiness checklist.

Ready to ship your app?

Ployed handles deployment, security and monitoring. A real engineer reviews every release. You keep building in the canvas editor.

Start your project