## The short version

The best deployment approach for a vibe-coded app is the simplest one that supports its real runtime, data, and release needs. Use a repo-first deployment agent when you own the source, want to keep using your current coding tool, and need repeatable releases without assembling cloud infrastructure. Use another approach when the app requires unsupported infrastructure or must stay inside a specific builder.

## WHO THIS IS FOR

This guide is for AI-native builders deciding how to move a working prototype into a shareable and repeatable production deployment.

## STEP BY STEP

### How to do it

1. **Confirm source access**  
   A repo-first deploy works best when you can open the full application source and production build locally. Builder-only projects may need an export or the builder's own hosting.

2. **List runtime and data needs**  
   Identify the language, web processes, workers, databases, volumes, object storage, environment variables, and custom domains the app actually uses.

3. **Check repeat release needs**  
   If the app will keep changing, prefer a workflow that records configuration, detects unchanged releases, preserves durable resources, and exposes failure logs.

4. **Match control to the project**  
   Use an application platform when deployment speed and operational simplicity matter most. Use lower-level infrastructure when hardware, networking, orchestration, or compliance control dominates.

5. **Test with a real release**  
   Deploy the actual app, not a toy substitute. Verify build, startup, a public route, data persistence, environment variables, and a second release.

## FIT CHECK

### What this supports

#### Good fit
- Source-controlled AI-built applications
- Agent-driven planning and deployment
- Repeat releases with visible diagnostics
- A path from generated URL to custom domain

#### Know the limits
- No one platform is the best fit for every runtime or compliance requirement
- A prototype may need production fixes before any platform can run it safely
- Generated code still requires security, data, and behavior review

## TROUBLESHOOTING

### Common failure modes

#### Choosing from marketing claims alone
Run the real app and verify the complete first and second deployment workflows.

#### Ignoring durable data
Identify every runtime write and move required state to a volume, database, or object store.

#### Treating the coding agent as an unlimited operator
Use scoped credentials, budget controls, reviewable plans, and clear deployment contracts.