On this page
Compute, storage and egress
A Cloud workspace uses three things besides model credits. Compute and disk have their own monthly allowance in your plan; egress is included within fair use:
| Used when | Plan allowance (Pro) | |
|---|---|---|
| Compute hours | The workspace is running | 40 hours |
| Disk | The workspace exists, running or not | 5 GB in total |
| Egress | The workspace sends data out to the internet | Fair use |
Compute hours
One compute hour is one hour on the standard machine. Bigger machines use hours faster:
| Machine | Good for | Uses per hour of running |
|---|---|---|
| 2 shared CPUs · 4 GB | Most repositories | 1 hour |
| 2 CPUs · 4 GB (dedicated) | Steady builds | 3 hours |
| 4 CPUs · 8 GB (dedicated) | Large builds and test suites | 6 hours |
Time is counted by the second. A workspace that runs for 90 seconds on the standard machine uses 90 seconds, not a whole hour. Pro's 40 hours are 40 hours on the standard machine, or 6 hours 40 minutes on the largest one.
Only running time counts. A workspace that is asleep or stopped uses no compute, and workspaces go to sleep on their own after 10 minutes without activity. Change that in Workspace settings → Machine → Idle timeout (minutes).
While a workspace runs, a small amount of compute is set aside ahead of time. It shows as reserved in Billing and turns into real usage as the time is counted.
If your compute hours run out while a workspace is running, Cloud stops it after a short grace period so that current work can finish. Its files are kept. Start it again after a top-up with Overage on, or after your monthly reset.
Disk
Each workspace has a persistent disk that holds your repository, dependencies and chats. A new workspace's disk is 5 GB unless you choose another size (1 to 500 GB) under Advanced. A disk counts at its full size for as long as the workspace exists, even while the workspace sleeps or is stopped. That is what lets a workspace wake up exactly where it left off.
Your plan's disk allowance covers the total size of all your workspaces' disks. On Pro, whose allowance is 5 GB, one workspace with the default disk fills it. If your disks fit, disk costs nothing extra. Beyond it, with Overage on, extra disk costs 500 credits per GB per month ($0.50), counted by the second:
| Extra disk | For | Credits |
|---|---|---|
| 1 GB | 30 days | 500 |
| 1 GB | 1 day | about 17 |
| 10 GB | 30 days | 5,000 |
With Overage off, Cloud does not create a workspace whose disk would go past the allowance. If your disks are already over it, for example after you move to a smaller plan, and purchased credits cannot pay for the extra, your running workspaces are stopped and you get an email. Their files are kept: delete a workspace or add credits with Overage on to run them again.
To stop paying for a disk, delete the workspace. Keep a copy of anything you need first.
Egress
Egress is data your workspaces send out to the internet, such as a git push or an upload. Data coming in, like a
git clone or npm install, is free, as is traffic between your workspace and Haarness itself.
Every plan includes ample egress for normal development work, so pushes, deploys and downloads of your own results never run into it. Only unusually large transfers beyond that fair-use amount are charged, and only with Overage on, at 100 credits per GB ($0.10).
Overage rates
When Overage is on, purchased credits pay for usage beyond your plan at these rates:
| Credits | In dollars | |
|---|---|---|
| Compute, standard machine | 120 per hour | $0.12 per hour |
| Disk | 500 per GB per month | $0.50 per GB per month |
| Egress beyond fair use | 100 per GB | $0.10 per GB |
Larger machines cost 3 or 6 times the compute rate. With Overage off, running out of an allowance stops the work instead of charging anything.
Keep the bill small
- Let workspaces sleep. Keep the idle timeout short and close terminals you no longer need.
- Use the standard machine unless a build really needs dedicated cores. It is 3 to 6 times cheaper.
- Pick the disk you need. The disk size is fixed when you create a workspace, so start new workspaces smaller if you can.
- Delete workspaces you have finished with, after you push or download anything you need.
- Set spending caps on API keys that call models from scripts.