Start with included runtime, scale up when a session needs more power, and keep heavy GPU work as a clear pay-as-you-go add-on.
Pick a monthly plan for included runtime and sandbox limits. You can manage your plan from the dashboard.
Try Trifecta Cloud before adding a payment method.
10 Launch-hours included
Low-cost hosted sandboxes for occasional personal agent work.
60 Launch-hours included
More runtime and concurrency for regular cloud development.
150 Launch-hours included
Shared hosted capacity for small teams and heavier agent sessions.
400 Launch-hours included
Customers can use included Launch-hour credits or continue pay-as-you-go after credits are used. Larger sandboxes consume credits in proportion to their cost.
1 vCPU / 2 GiB RAM / 10 GiB disk
$0.12/hr
Light agent sessions, reviews, and small edits
2 vCPU / 4 GiB RAM / 10 GiB disk
$0.24/hr
Package installs, test runs, and everyday coding
4 vCPU / 8 GiB RAM / 10 GiB disk
$0.48/hr
Heavier builds, parallel work, and larger repositories
GPU sandboxes are planned as pay-as-you-go add-ons for higher tiers, billed only while the GPU sandbox is running.
Pro and Team plans
$4.75/hr
Billed only while the GPU sandbox is running
Pro and Team plans
$5.95/hr
Billed only while the GPU sandbox is running
Trifecta Cloud provides the computer. It does not resell paid model subscriptions or hide token costs inside the sandbox price.
Sign in before creating or managing cloud sandboxes.
Included runtime is measured in Launch-hour credits. Larger CPU sandboxes burn credits faster.
Stopped sandboxes keep their filesystem but do not consume CPU/RAM runtime.
Each sandbox includes 10 GiB disk. Additional persistent storage is planned at $0.12/GiB-month.
The sandbox image will include supported AI tooling and OpenCode integration.
OpenCode free model options will be available in the sandbox image where supported.
Users supply their own subscription auth or API keys for paid model services and CLIs.
Included credits, idle auto-stop, and metered overages keep cloud usage predictable without asking users to commit to large infrastructure contracts.