Cloud platforms (one-click deploy)

TL;DR

Each platform offers a one-click deploy button that reads the repo’s deploy.json and provisions the services automatically. You still need BYOK keys — paste them into the provisioned dashboard’s env-var section.

Platform	Cost (starter)	Cold-start	Best for
Railway	~$10/mo	instant	Solo operators, small teams
Render	~$15/mo	instant	Moderate scale, clean TLS
Fly.io	~$8/mo (3 machines)	~2s	Global, multi-region
DigitalOcean App Platform	~$15/mo	instant	Teams already on DO droplets

All of them work. Ignore the marketing copy — pick based on which dashboard you already know.

Railway

1. Click Deploy to Railway.
2. Sign in. Railway provisions:
   • backend (Node)
   • frontend (Node)
   • postgres (managed)
   • redis (managed)
3. The template pre-fills env vars. You only need to add your BYOK keys:
   • ANTHROPIC_API_KEY or OPENAI_API_KEY or GEMINI_API_KEY
4. Deploy. Open the generated URL.
5. You’re in the setup wizard.

Production-ready Railway

• Custom domain. Railway → Settings → Domains → add yours. Automatic TLS via Let’s Encrypt.
• Postgres backups. Railway backs up nightly; configurable to hourly on paid plans.
• Logs. Railway’s log viewer is usable. For long retention, pipe to Logtail / Axiom / Better Stack.

Render

1. Deploy to Render.
2. Render reads render.yaml and provisions services.
3. Add BYOK keys in the Render dashboard (one service at a time).
4. First deploy takes ~5 min; subsequent deploys ~2 min.

Render specifics

• Render Postgres has a free tier but it’s deleted after 90 days of inactivity. Upgrade for anything real.
• Static frontend option. You can deploy the frontend as a Render static site (cheaper) — the template does this by default.
• Background workers (queue consumers) share the backend service; Render doesn’t need a separate worker.

Fly.io

1. Install flyctl.

2. git clone https://github.com/workforce0/workforce0
   cd workforce0
   fly launch

3. fly launch reads fly.toml, asks confirmation, provisions:
   • Backend app (3 machines across regions)
   • Frontend app
   • Postgres cluster (fly pg create)
   • Upstash Redis (fly ext redis create)
4. Set BYOK keys:

   fly secrets set ANTHROPIC_API_KEY=...
   fly secrets set JWT_SECRET=$(openssl rand -hex 32)

5. fly deploy.

Fly specifics

• Multi-region by default. Backend scales to zero when idle; scales up on the first request. ~2 second cold-start.
• Upstash Redis is single-region; the backend’s BullMQ setup handles the cross-region latency fine at <100ms.
• Volumes for uploads. fly volumes create workforce0_uploads and mount in fly.toml.

DigitalOcean App Platform

1. Deploy to DO.
2. App Platform reads .do/app.yaml and provisions components.
3. Add BYOK keys in the component settings.
4. Deploy.

DO specifics

• Managed Postgres add-on — $15/mo, minimum. DO’s DATABASE_URL has SSL mandatory; already handled.
• Managed Redis — $15/mo additional. Cheaper to run Redis as a worker component on a $5 droplet for hobby.
• Nice TLS story — *.ondigitalocean.app has automatic TLS, with your own domain just a CNAME.

What you can’t do on one-click platforms

• Run the agent daemon for code-gen — it needs your local CLI subscription. Leave agent disabled.
• Truly high-scale workloads — once you’re over 100 concurrent briefs, move to Kubernetes.
• Network-isolate Postgres — PaaS Postgres is always reachable over the internet with credentials. Fine for most; not for compliance-heavy workloads.

Migrating between platforms

All four platforms expose a Postgres connection string. To migrate:

# On the old platform
pg_dump -Fc -d $OLD_DATABASE_URL > dump

# On the new platform
pg_restore -d $NEW_DATABASE_URL dump

Point the new platform’s backend at the restored DB, deploy, done. Zero data loss as long as you pause the old deployment during the dump.

Costs

Rough monthly bills at typical usage (100 briefs / month):

Platform	App tier	Postgres	Redis	Total
Railway	$10	included	included	~$10
Render	$7	$7	$10	~$24
Fly.io	$8	$0 (dev)	$10 (Upstash)	~$18
DO	$12	$15	$15	~$42

BYOK API costs are additional and depend on your model mix — see Cost caps.