> ## Documentation Index
> Fetch the complete documentation index at: https://vetta.sh/docs/llms.txt
> Use this file to discover all available pages before exploring further.
# Limits & billing
> Concurrency caps, runtime ceilings, output limits, and the exact metering model for compute.
Compute is metered from the seconds a computer actually ran, booked against your [budget](/docs/concepts/budgets) at the end of each turn and drawn from the organization's real-USD balance. This page is the reference for the caps and the exact per-unit rates.
Every rate on this page is **the price you pay** — the number booked to the [ledger](/docs/platform/billing) and reported on your spend breakdown. Model inference is priced separately; see [Model router & inference](/docs/concepts/model-router).
## The compute meter
While a computer is **running**, it is metered **per second** on its *provisioned* resources — vCPU and memory. There is **no creation fee**, and **disk is not metered**: the boot disk is inside the allowance the computer ships with. Billing granularity is per-second.
| Resource | Per second | Per hour |
| ---------------- | ------------ | --------- |
| vCPU | `$0.000014` | `$0.0504` |
| Memory (per GiB) | `$0.0000045` | `$0.0162` |
| State | Billed |
| ------------------- | ---------------------------------------------------- |
| `running` | Per-second vCPU + memory. |
| `sleeping` / paused | **Nothing** — no vCPU, no memory, no storage charge. |
| `destroyed` | Nothing further. |
A session's computer is **`1 vCPU / 1 GiB`**, so a full hour of *running* time is **`$0.0504 + $0.0162 = $0.0666`**. Because idle maps onto the [runtime's](/docs/concepts/runtime) wake/sleep cycle, a computer waiting between turns is metered for nothing at all — the mechanism behind Vetta's low cost on multi-hour tasks.
## The browser
The optional [browser](/docs/computer/browser) carries **no charge of its own**. Browser time is not metered, there is no browser line on the [ledger](/docs/platform/billing) and no `browser` component on your spend breakdown — what a browsing agent pays for is the tokens it spends driving the browser and the computer time of the session driving it.
Its limits are still real: see the per-session timeout and the concurrency cap below.
## Concurrency
Every computer counts toward your organization's concurrency cap the entire time it is not `destroyed` — `running`, `sleeping`, and `parked` all occupy a slot; only a destroy frees one. The defaults below are per organization and are **raised on request** — they are safety rails, not hard product ceilings.
| Concurrency limit | Default | Notes |
| ----------------------------------- | ------- | -------------------------------------------------------------------------- |
| Concurrent computers per org | `25` | Counts every non-`destroyed` computer (`running` + `sleeping` + `parked`). |
| Concurrent browser sessions per org | `10` | Across all computers. See [Browser](/docs/computer/browser). |
## Boot & wake latency
Only cold boots pay the full provisioning cost; the wake/sleep cycle the [runtime](/docs/concepts/runtime) uses between turns is near-instant, which is what makes idle cheap.
| Transition | Typical latency | What happens |
| ------------------------------------------------------------------- | --------------- | ------------------------------------------------------------------- |
| Cold boot (fresh `create`) | \~2 s | A micro-VM is provisioned from a clean image. |
| Boot from a [snapshot](/docs/computer/snapshots-and-volumes#checkpoints) | \~3 s | Disk state is restored before the VM runs. |
| Wake from `sleeping` | \~300 ms | Disk and memory are preserved; the next exec or file op resumes it. |
| Resume from `parked` | \~300 ms | Same as a wake, but requires an explicit `resume`. |
The fully virtualized **isolate** mode (`mode: "isolate"`, **coming soon**) targets near-instant boot and near-zero idle; the numbers above describe the micro-VM that ships today.
## Limits
| Limit | Value | Notes |
| --------------------------------- | ----------------------------------------------- | ----------------------------------------------------------- |
| Max runtime | `max_duration_hours` per computer (default 12h) | A computer exceeding its ceiling is auto-destroyed. |
| Idle auto-pause | `idle_timeout_minutes` per computer | Optional; pauses to storage-only after idle. |
| Auto-destroy on credit exhaustion | balance ≤ 0 | Protects against runaway spend. |
| Shell output | 262,144 chars each for stdout/stderr | See [Shell](/docs/computer/shell). |
| File read/write | 4 MiB per file | See [Filesystem](/docs/computer/filesystem). |
| Persistent volume size | 1 MiB – 100 GiB | See [Snapshots & volumes](/docs/computer/snapshots-and-volumes). |
| Browser session timeout | 1 min – 6 h (`timeout_minutes`, max 360) | Hard per-session TTL. |
`max_duration_hours` bounds a computer's **total lifetime from create to destroy** — it is wall-clock, not billed compute-seconds, so a computer that sleeps most of that window still hits the ceiling on schedule. Only `running` seconds are metered; the ceiling is a safety cap on how long the computer may exist.
A computer is **auto-destroyed** when it exceeds its max runtime or the organization's balance reaches zero. Persist anything you need to keep with [`publish_file`](/docs/capabilities/files) or a [volume](/docs/computer/snapshots-and-volumes) before then.
## Inspecting spend
Compute spend appears as the `computer` component of the [agent's spend breakdown](/docs/concepts/budgets) and as a line item on the organization [ledger](/docs/platform/billing).
```bash CLI theme={"system"}
vetta agent spend Refunder --by component
```
Spend is reported on the wire as integer **micro-USD** (`1 USD = 1_000_000`); the CLI renders dollars for humans.
```json theme={"system"}
{ "by_component": { "model": 11902000, "computer": 481000 } }
```
The full computer surface at a glance.