Skip to content

Spend and caps

View as Markdown

How Guv bills in pounds, per second of actual use, and how hard spend caps stop surprise bills.

Guv bills in pounds sterling. Your plan includes some usage each month, and anything beyond that is charged per second on what your services actually use. Builds are free. See pricing for the current rates, and use the calculator there to estimate a bill.

Resource How it is measured
CPU vCPU-seconds actually used. A service that idles at 2% of a vCPU pays for 2%.
Memory Your working set, sampled every 10 seconds. Cache the kernel can drop is not counted.
Disk Space used on each volume, sampled hourly and averaged over the month.
Outbound data Bytes leaving the platform. 100 GB a month is included.

Every workspace has a monthly spend cap from the moment it is created: £10 on Hobby and £50 on Pro. You can change it, but Guv never raises it for you, and an agent can never raise it at all.

  • At 50% and 80% you get an email, a banner in the dashboard and a webhook.
  • At 100%, the on_cap action runs. The default, sleep-non-production, puts staging and preview environments to sleep and keeps production serving.

You can set the cap and its behaviour in guv.yaml:

spend:
monthly_cap: 400GBP
on_cap: sleep-non-production # or pause-all, or notify-only
warn_at: [50, 80]

notify-only never pauses anything. Because that removes the protection, it needs a payment method on file and acknowledge_no_pause: true.

The dashboard and guv spend --forecast project your month-end bill from your usage so far, with a short explanation for each service, such as “api used 0.3 vCPU on average”.

Agent tokens have their own daily budget for how much they may increase spend. A change that would go over it returns agent_budget_exceeded and asks a human.

Prices are shown without VAT for businesses and with VAT for consumers. Invoices are VAT-correct. Guv uses Stripe to take payments.