Platform beta The version-controlled operations layer

Run your servers like
you run code.

Branch, diff, tag and roll back live servers. One panel for every host you run.

Free for the core toolsBring VMs from any providerCancel anytime
apt upgradenginx.conf editknown-goodcf branch · test the upgrademergebad deploycf checkout known-goodprod
apt upgradenginx.conf editknown-goodcf branchmergebad deploy↺ cf checkout known-goodprod

Bring servers from any provider

ComputeForge cloudDigitalOceanHetznerAWSLinodeYour homelab
01Our cloud

git branch, but for the running server.

A running copy of a ComputeForge server on its own firewalled address, in a minute or two. Break the copy, not prod. Servers at other providers: Coming soon

How it works
isolated · ssh from your addresspg17-testprod-dbcf diffbreak the copy, not prod
02Any provider

See what changed, when, and who did it.

History, diff and blame for every tracked file on the box, in the panel and the API.

How it works
c8d2e1apt upgrade · 47 packagesb1f7a3nginx/api.conf · alice9e4f02restart nginx · alice7d3c0asshd_config · bob/etc/nginx/api.conf- proxy_pass …/v1;+ proxy_pass …/v2;blame: alice@laptop · 19:22who · what · when
03Any provider

Tag the known-good state. Roll back in one click.

Restore a tag's files on any connected server that allows remote changes, with no downtime. A whole server comes back on our cloud, same IP.

How it works
before-pg17deploy brokecf checkout --tag before-pg17files: no downtimewhole server: same IP
04Any providerComing soon

git push, but it's your VM moving providers.

Copy a running server to another provider on the same Linux release, cut over when ready, roll back for 24 hours.

How it works
provider Aprovider Bprod-webprod-webfull copy · delta copyrollback · 24 hDNS cutover when you say go
The panel Live today for ComputeForge cloud servers

One panel for your servers.

Servers, volumes, snapshots, networking and access, in one place.

The ComputeForge panel's Servers page, light theme: server list with status, CPU, memory, disk and address. Demo data.The ComputeForge panel's Servers page, dark theme. Demo data.
Demo data. The panel follows your light or dark choice too.
05Any provider

DNS that fails over when a probe goes red.

Lifeguard probes any VM; Forge DNS swaps the record when it falls and notifies you.

How it works
Forge DNSapp.example.comswappedweb-1probe failedweb-2healthyLifeguard probeSlack · Discord · ntfy
06Any providerComing soon

Stop being on call. Start being protected.

Spot an anomaly, scan, propose the fix. You decide how much it may do on its own.

How it works
detectscanproposeactfail over · branch· check outyou choose how far it goesalert-onlypropose+approvefull-autonomousevery step lands in cf log
07Any providerComing soon

Describe your stack once.

Services, dependencies, standbys and autonomy in one YAML file, or in plain English.

How it works
prod-web ×3standbyprod-dbprod-redispropose+approvedb: never auto
08Any provider · laptops · homelab

Every server on one private network.

One click puts a connected server on your WireGuard mesh. Laptops and homelab boxes can join too.

How it works
web-1our clouddb-1our cloudedge-3another providerlaptopyounashomelabWireGuard
09Any providerComing soon

AI that reads the diff.

A plain-English reading of what changed, and a suggested fix you can apply or ignore.

How it works
10Any provider

Backups in a format you own.

A checkpoint into your own Borg or Restic repository, with a passphrase only you keep. DNS zones as BIND, mesh as plain WireGuard.

How it works
checkpointsexport · open formatsBorgcheckpointsResticcheckpointsBINDDNS zonesWireGuardmesh configyour key · restore on any Linux box
11Any provider · out-of-band on our cloud

Stop trusting the OS to tell you the truth.

A scan checks the disk against what the OS reports and against package hashes, to catch hidden files and swapped binaries.

How it works
what the OS says/sbin/sshd✓/etc/cron.d✓/tmp✓what the disk says/sbin/sshdhash ≠/etc/cron.dmodified/tmp/.xhiddenchecked against the diskanomalies found · capsule sealed
Features More than version control

Change the server you have. Don't rebuild it.

More RAM, a bigger disk or a different OS should not mean starting over. These are the tools for changing the server you already run; the ones marked “Coming soon” are built but not open to every account yet.

Archive a server, keep it for later

Done with a server but not ready to lose it? Archive it into protected storage, download it as a standard disk image, and restore it when you need it.

Old branch snapshots into object storage

Promoted a branch? Move the snapshots it was made from into object storage. The copy is checked in full before the snapshots are removed, then sits under Archive, billed at the object storage rate.

Backups in a format you own

Export a tagged checkpoint into your own Borg repository, encrypted with a passphrase only you keep. Restore it on any Linux box.

Know before your customers do

Health checks on any server, alerts by e-mail, Slack, Discord, ntfy or webhook, and DNS that fails over when a check goes red.

Checkpoints and branches

Tag a known-good state and roll back to it. Branch a running copy of a server on our cloud and try the risky change there first.

Scan the disk from outside the server

Rootkits lie to the OS. On our cloud the integrity scan reads a copy of the disk from outside the server and checks its files without trusting what the OS reports.

Pay by the second

Delete a test server after ten minutes and pay for ten minutes (at least one cent). Or pay the monthly rate for a server that runs all month.

More CPU and RAM, no restartComing soon

A traffic spike should not mean a reboot. Add CPU and RAM to a running server, take them back off when it is quiet, and pay for the extra by the second while it is there.

RAM that follows the loadComing soon

Set a floor, a ceiling and a spending limit. The server gets more RAM when it runs short and gives it back when the load drops.

Grow the disk while it runsComing soon

Disk full? Make it bigger without stopping the server. The panel shows how to grow the file system: at the next restart, or right away with one command.

Shrink the disk to a smaller planComing soon

Paying for disk you never use? Move to a smaller plan without reinstalling. Your data comes along, and you can undo it if something looks wrong.

Boot your own ISOComing soon

Need an OS we do not offer, or your own rescue tools? Boot a server from an ISO, from a web address, an upload or your own bucket, then detach it.

AMD EPYC processor
ComputeForge cloud

Host with us when the verbs need our metal.

AMD EPYC serversNVMe storage1 IPv4 + free IPv6Monthly or per-second

Per-second billing

Delete a CI runner after 17 seconds, pay for 17 seconds (at least one cent).

A person to talk to

A question about a charge or a stopped server goes to a real person, through a support ticket in the panel. Your activity log is private to you and your team.

Open-format export

Export a checkpoint into your own borg repository over SSH, encrypted with repokey-blake2 and a passphrase only you keep. Download any DNS zone as a standard BIND zone file. Export your mesh as plain WireGuard configs: we never hold your private keys. Your data, your format.

Compare Platform vs. hosting

The verbs across providers.

The verbs run over a small agent you install with one command from the panel. Full comparison →

Platform tierWorks on any provider you connect — DO, Hetzner, Linode, AWS, homelab, or our metal
CapabilityComputeForgeDO / Hetzner / LinodeDatadog / PagerDutyPulumi / Terraform
Change history for any server✓—partial (events only)—
Diff between two servers✓———
Branch a running serverComing soon———
Tag + restore✓partial (snapshots)—partial (state files)
Health-check-driven DNS failover✓partial (LB health checks)partial (alerts only)—
Cross-provider mesh✓———
cf push between providersComing soon———
Hosting tierOnly on ComputeForge servers
CapabilityComputeForgeDOHetznerLinodeAWS
Per-second billing (servers, volumes, snapshots, IPs)✓partial (Droplets only, 60 s min)——partial (60s min)
Pricing The cf CLI is free. Hosting tier priced below.

Pick a plan. Pay monthly or by the second.

ComputeForge cloud servers, in US dollars before tax. Per-second billing costs about 20 % more than the monthly rate.

Cloud 1 GB

Side projects & static sites
$5/mo
  • 1 GB memory
  • 1 vCPU AMD EPYC
  • 25 GB NVMe
  • 1 TB bandwidth / month
  • 1 IPv4 + free IPv6
Choose Cloud 1 GB

Cloud 64 GB

Heavy workloads
$227/mo
  • 64 GB memory
  • 20 vCPU AMD EPYC
  • 1.6 TB NVMe
  • 7 TB bandwidth / month
  • 1 IPv4 + free IPv6
Choose Cloud 64 GB

Cloud 128 GB

Database & enterprise
$477/mo
  • 128 GB memory
  • 32 vCPU AMD EPYC
  • 3.2 TB NVMe
  • 9 TB bandwidth / month
  • 1 IPv4 + free IPv6
Choose Cloud 128 GB

Every plan, add-ons and custom plans

Run your servers like
you run code.

Sign up, add a card, create a server. Delete it any time.