Run your whole stack from one file.

Services, workers, cron and databases — declared once, reviewed like code, friendly to the tools that write code for you.

Request access

Deploys.

Rolling and health-checked; one action rolls back. Push to deploy, median build 41 seconds.

Databases.

Postgres and Redis with nightly backups, point-in-time recovery, and restores rehearsed weekly on a throwaway node.

One file.

duploy.toml is the source of truth. No dashboard state to drift; every change from panel or CLI becomes a pull request.

Domains & TLS.

DNS checked automatically; certificates issue and renew themselves. Nothing to click through.

One log stream.

Builds, deploys, cron runs, backups and node events in one place — panel and CLI show the same stream.

Your agent already knows the CLI

duploy skill writes a skill file that teaches coding agents the whole command set. Every command takes --json, and plans are dry-run by default — an agent stages the change, you approve the pull request.

AI writes the change. You stay the reviewer.

Everything, at a glance

The full set. Defaults, not add-ons.

builds for Node, Python, Go, Ruby, Rust
Dockerfile respected when present
cron jobs with run history
persistent volumes
write-only secrets
instances reschedule on node failure
import from Procfile, render.yaml, fly.toml
export everything, no egress fee
review rules per environment
per-second billing, visible per node
Frankfurt + Ashburn, databases in-region
99.98% measured uptime, 90 days

Pricing

Solo$0 during beta · then $5/mo incl. usage
2 projects · pg + redis · nightly backups · custom domains
Team$20 / member / mo + usage
unlimited projects · PITR 7d · restore rehearsals · review rules
Request access →
metered rates
compute$0.021 / vCPU·h
memory$0.010 / GB·h
volumes$0.15 / GB·mo
egress$0.05 / GB · ingress free
Billed per second while running. Stopped services cost storage only.

Questions we actually get

Do I need a Dockerfile?

No. Builds detect the stack from the repository; a Dockerfile wins if present.

Where does my app run?

On machines we own in Frankfurt (fra1) and Ashburn (iad1) — not resold cloud. Databases stay in-region.

What can an agent actually break?

Nothing directly. Agents work through the same staged changes as humans — dry-run plans, then a pull request you approve. Production only moves on merge.

Can I take my data and leave?

duploy export hands you the file, volume archives and pg dumps any time. No egress fee on the way out.

$ duploy up
web · worker · pg — healthy in 48s
Request access