hostfleet Find my setup
From generated code to a real product

Your AI app works.
Where should it live?

Do not move the whole prototype into one mystery box. Split it into a fast frontend, a backend for secrets and slow work, and persistent data.

Map my app →
prototype.app
Built with AIYour working demo
!
A demo is not an architectureSecrets, long jobs, data, domains, and limits still need a home.
The production map

Three layers. Three jobs.

Most generated apps become easier to host once the code is separated by what it actually does.

Layer 1

Frontend

Pages, interface, static assets, and fast request handlers close to users.

Good homes
Cloudflare Pages · Vercel · Netlify
Keep out
Secrets · long jobs · state
Choose a frontend host →
API
Layer 2

Backend

Authentication, model calls, webhooks, queues, cron, agents, and slow work.

Good homes
Railway · Render · Fly.io · VPS
Needs
Logs · retries · secrets · uptime
Choose a backend →
SQL
Layer 3

Data

Users, app state, files, embeddings, queues, and the information that must survive deploys.

Good homes
Managed Postgres · object storage
Needs
Backups · migrations · access control
Plan the data layer →
Pick by maturity

How far should you move?

Migration is not a badge of seriousness. Keep the builder when it works; export only when a concrete limit justifies the operational load.

1Fastest

Stay in the builder

Use the included hosting and database for validation.

Best while:the limits are not hurting you
2Balanced

Export the frontend

Own the domain and web deployment; keep managed services.

Best when:you need portability or custom CI
3Most control

Own the full stack

Operate web, workers, state, monitoring, and recovery.

Best when:the product can pay for operations
Before launch

What usually breaks first?

Generated code can look complete while the production behavior remains untested.

01Secrets in the browserAPI keys and privileged calls belong behind a server boundary.
02Long jobs time outMove generation, scraping, and workflows into a durable worker.
03No retry pathModel and webhook failures need idempotency and recovery.
04Data disappearsEphemeral filesystems are not databases or backups.
05No useful logsCapture request IDs, job state, failures, and provider latency.
06$Spend has no ceilingAdd rate limits, quotas, budgets, and abuse controls.
Use the complete production checklistOpen 12 failure modes →
Deep research

Deployment research library

Detailed platform limits, comparisons, migration instructions, and backend patterns live here when you need the evidence behind the map.

deploy ai apps Cloudflare Workers + Railway for AI webhooks (July 2026): keep ingress thin and move slow jobs to a worker A source-backed deployment guide for handling AI webhooks with Cloudflare Workers at the public edge and Railway for durable job processing, retries, private services, and slow agent work. 2026-07-22 · Read the evidence → deploy ai apps Cloudflare Pages for AI-generated frontends (July 2026): the cheapest clean default if the logic stays thin A July 2026 HostFleet deployment guide to Cloudflare Pages for AI-generated frontends, focused on build caps, the $5 Workers Paid step, and where the platform stops fitting. 2026-07-15 · Read the evidence → deploy ai apps Netlify for AI-generated frontends (July 2026): flexible, but the credit meter changes the math A July 2026 HostFleet deployment guide to Netlify for AI-generated frontends, focused on credit-based pricing, deploy-preview workflow, Next.js support, and the pause risk when monthly credits run out. 2026-07-14 · Read the evidence → deploy ai apps Vercel for AI-generated frontends (July 2026): best preview UX, paid the moment it is real A July 2026 HostFleet deployment guide to Vercel for AI-generated frontends, focused on Hobby's non-commercial limits, the real Pro floor, and the function/runtime ceilings that shape light AI features. 2026-07-14 · Read the evidence → deploy ai apps Best hosting for AI cron jobs (July 2026): Cloudflare vs Render vs Railway vs one cheap VPS A July 2026 HostFleet guide to the best hosting for AI cron jobs, focused on short scheduled API work, managed cron containers, and when a cheap VPS is the more honest answer. 2026-07-10 · Read the evidence → deploy ai apps Coolify vs Railway vs Fly.io vs Render for small AI apps (July 2026): when self-hosting wins and when managed platforms are worth it A July 2026 HostFleet comparison of Coolify vs Railway vs Fly.io vs Render for small AI apps, focused on the real cost shape of one app plus one worker and the trade between self-hosting and managed platforms. 2026-07-10 · Read the evidence → deploy ai apps Fly.io for AI backends and workers (July 2026): where the cheap Machine fits and where volumes bite A July 2026 HostFleet guide to Fly.io for AI backends and workers, focused on the real floor for always-on CPU services, where autostop actually helps, and why local volumes are the platform's sharp edge. 2026-07-10 · Read the evidence → deploy ai apps Where to host LiteLLM for small teams (July 2026): managed proxy vs one small VPS A July 2026 HostFleet deployment guide to hosting LiteLLM for small teams, including when to skip the proxy, when Railway is the cleanest managed home, and when a small VPS is the better fit. 2026-07-10 · Read the evidence → deploy ai apps Where to host MCP servers for small teams (May 2026) Where to host MCP servers for small teams in 2026: keep personal stdio servers local, use Railway for the default managed case, Cloudflare for thin HTTP tools, and Fly.io or Hetzner when you need always-on control. 2026-07-10 · Read the evidence →
Show 10 more research notes