Hosted machines and billing
Hosted machines are available on any plan when hosted availability is enabled. Availability is currently disabled pending production release verification. Flowviant uses Amazon Web Services EC2 and encrypted EBS disks in the machine’s configured AWS region. Hosted Claude and Codex use Flowviant’s managed Anthropic and OpenAI API accounts.
Prices and credits
In Account → Billing → Hosted spending, read the published rate schedule, remaining credits, reserved credits and available credits. Each $1 paid adds $1 of credit. AWS compute, disk storage and related usage cost the published regional base price plus 50%; managed AI costs the selected model’s published price plus 10%. The rate schedule shows actual configured numeric rates, region/model, effective date and cloud price source. No default or example price is charged. Prices must be verified and published there before provisioning is enabled.
Credit is reserved before paid work begins. Reservations include seven days of disk storage and a bounded compute interval; model requests reserve their maximum priced usage. Measured usage is charged once, and unused credit is released after reconciliation. An uncertain provider response keeps its reservation until usage is established. A top-up return page alone does not prove payment: use Check payment to reconcile the saved payment. A failed checkout can be reopened without creating another payment.
The spend cap
Set the monthly spend cap in Hosted spending. It is shared by hosted machines paid for by this account across projects and resets on the first day of each month in UTC. Pending reservations count against it. Reaching the cap or running out of available credit pauses new hosted work and stops compute after the daemon drains. A top-up or raised cap does not start a machine automatically.
Provision, stop and resume
The project’s owner provisions from Project settings → Machines after installing and authorizing the hosted GitHub App for the selected repository. Project members can read the state; its paying account controls paid lifecycle changes. The server distinguishes provisioning, running, stopping, stopped, limit-paused and failed states. A failed operation keeps its error and retry door.
Ten minutes of confirmed idle activity stops compute automatically. Active agent work or deployment prevents idle shutdown. Stop drains work and preserves the checkout. Once the server measures stopped, Start or Resume checks funding and the cap before starting the same disk and repository again.
Disk retention and deletion
Stopped disks still incur storage charges. The approved release policy reserves seven days of disk storage before starting compute. When funding ends, the disk is retained for seven days unless funded again, with a warning at pause and 24 hours before permanent deletion. Automatic retention warnings and deletion are release prerequisites and are not enabled yet.
Delete permanently removes the hosted disk, including unmerged work. Merge or copy any work you need first. Existing merged changes in GitHub and project board records are separate from the hosted disk. See the privacy policy.