Monthly subscriptions with credits included. Add more anytime with credit packs. Pay only for what you use.
Every plan includes credits for compute, storage, and operations. Paid tiers add a monthly credit allowance; Free includes a one-time starter balance.
Custom credit volumes, dedicated concurrency, reserved capacity, and contractual SLAs. For large-scale and regulated workloads.
All prices in EUR. Unused subscription credits roll over for one month. Cancel or change plans anytime.
Top up anytime. Purchased credits are added to your org pool alongside subscription credits and allocated to environments as needed.
Credits are the single unit across all KernStack services. No per-service billing, no surprises.
Your subscription credits are allocated at the start of each billing month and shared across all services — Functions, Workers, Tables, Buckets, Inference, Training, and Search.
Unused subscription credits roll over for one billing period. This applies to your monthly allocation only — credit pack top-ups add to your balance separately and are not subject to the rollover cap.
Credits deduct automatically from each environment's balance as you use the platform. Usage maps to compute, storage, and network — see the Price Reference below for rates and how each dimension is measured.
If your balance reaches zero, new requests are throttled — not terminated. Upgrade, top up, or wait for the next billing cycle. Your data is never deleted.
We map billable work across the stack to the same three primitives — compute, storage, and network — so usage is expressed in a single internal model that mirrors real resources instead of a separate fee per product surface.
| Dimension | Rate (credits) |
|---|---|
| Compute | 0.01 / GB-second |
| Storage (Tables, Buckets, Search index storage, etc.) | 0.04 / GB-hour |
| Network egress | 9.00 / GiB |
For duration-metered usage, your bill is compute + storage + network at the rates above, summed and rounded up to a whole credit. Compute uses (RAM in mebibytes ÷ 1024) × wall-clock seconds — GiB-seconds — × the compute rate; the table label “GB-second” is shorthand. Network is billed per GiB of API payload and response traffic (bytes ÷ 1024³) × the network rate; Functions, Workers, and Inference also bill outbound egress from guest code.
| Model | Unit | Batch | Credits |
|---|---|---|---|
text-embed-v1 |
Tokens | 1,000 | 3 |
sentiment-multilingual-v1 |
Tokens | 1,000 | 2 |
face-detect-v1 |
Images | 1 | 3 |
label-detect-v1 |
Images | 1 | 3 |
text-detect-v1 |
Images | 1 | 3 |
moderation-v1 |
Images | 1 | 3 |
image-quality-v1 |
Images | 1 | 3 |
Each catalog model charges in steps: count prompt_tokens in chunks of 1,000, or one step per images. Each step uses fixed compute and egress assumptions at the rates above (rounded up). Total = steps × the per-step credits in the table (e.g. 2,500 sentiment tokens → 3 steps × 2 = 6 credits).
Custom inference (non-catalog invoke) uses wall-clock time, endpoint memory, and response size as egress — same compute + egress formula as other metered APIs.
Rounded monthly examples mixing compute, storage, egress, search, and some catalog model usage. Real usage varies.
Typical REST + auth, profiles and feeds, light background jobs and push hooks.
Approx. total~1.9k cr
Predictable API and worker load, shared tables and search, with selective catalog models for text and vision.
Approx. total~8.1k cr
Higher RPS, heavier search and recommendations, bursts of generative media — often a subscription plus top-up packs.
Approx. total~28.7k cr
KernStack is currently in preview. Sign up to start building, or get in touch to learn more.
Get in Touch