---
title: "mrin — pricing"
description: "Billed for the work your machine does, not for the space it rents. We bill nothing for idle machines: running bills vCPU and memory, hibernated bills nothing. No base fee, storage and forks free. Indicative alpha pricing."
url: https://mrin.ai/pricing
---
# mrin pricing (indicative — the seeded catalogue, not a published price list)

The meter bills resident wall-clock hours at the size the VM was configured
with (vCPU-hours and GiB-hours); guest CPU utilisation never enters the bill.
Suspend parks the vCPUs but keeps RAM resident, so only memory bills.
Hibernate commits the machine to disk and stops both clocks. Snapshot storage,
forks and restores are metered but free. There are no base fees; the plans
differ only in concurrency cap and retention. Usage rates are flat across
plans. There is no free tier in the private alpha.

| Plan | Base fee | Concurrent sandboxes | Retention | vCPU-hour | GiB-hour | Storage | Fork / restore |
|---|---|---|---|---|---|---|---|
| Dev | none | 10 | 30 days | $0.04032 | $0.01296 | free | free |
| Pro | none | uncapped | 90 days | $0.04032 | $0.01296 | free | free |

Worked example: 7,300 sandbox hours a month (ten sandboxes, always alive) at
2 vCPU / 4 GiB with 15% CPU utilization, meaning the machine is resident for
15% of those hours and hibernated for the rest:
vCPU 2,190 vCPU-hours × $0.04032 = $88.30; memory 4,380 GiB-hours × $0.01296 =
$56.76; total about $145. The same fleet always running would be about $967.

The concurrency cap counts live branches; a hibernated sandbox still holds
its slot until released. The alpha is self-hosted with our help; hosted is in
progress. Real numbers come out of a call. Interactive calculator: https://mrin.ai/pricing

## Rate card

The same on every plan. Each meter bills per billable unit:

| Meter | Billable unit | Price |
|---|---|---|
| `vm.vcpu_us` | vcpu-hour | $0.04032 |
| `vm.mem_mib_us` | gib-hour | $0.01296 |
| `storage.blob_bytes` | gb-day | free |
| `fork.count` | operation | free |
| `restore.count` | operation | free |

## Plans

### Dev

- Base fee: none
- Concurrent sandboxes: 10
- Snapshot retention: 30 days
- Usage: the rate card above

### Pro

- Base fee: none
- Concurrent sandboxes: uncapped
- Snapshot retention: 90 days
- Usage: the rate card above

## What each sandbox state bills

The meter follows the branch state, not guest CPU load:

| State | vCPU-hours | GiB-hours | Notes |
|---|---|---|---|
| idle or running | billed | billed | Resident at its configured size, whatever the guest is doing |
| suspended | not billed | billed | vCPUs paused, memory stays resident; resume is instant |
| hibernated | not billed | not billed | Committed to disk and the VM freed; resume restores from the commit |
| released | not billed | not billed | VM stopped for good; its commits remain until retention expires |

Snapshot storage, forks and restores are metered and priced at zero on both plans.

## Choosing a plan

| | Dev | Pro |
|---|---|---|
| Base fee | none | none |
| Concurrent sandboxes | 10 | uncapped |
| Snapshot retention | 30 days | 90 days |
| Usage rates | rate card above | rate card above |
| Best for | Trying mrin on one workload, small agent fleets | Fleets above 10 live sandboxes, longer rollback windows |

## Included on every plan

- Every exec committed as a snapshot of the whole microVM, memory and disk
- Fork any commit into a new sandbox; restore any commit as a new branch
- Replay of each branch's terminal output, in order
- Suspend, hibernate, resume, and scale up a running VM
- The REST API, one multiplexed WebSocket, the `mrin` CLI, and the TypeScript, Python and Go SDKs
- Webhooks for commits, spawns, forks and branch lifecycle events
- API keys with viewer, developer, admin and owner roles

## Billing FAQ

**Does an idle sandbox cost money?** Yes, while it is resident. Hibernate it
and it costs nothing until you resume it.

**Are forks and restores charged?** They are metered but priced at zero, as is
snapshot storage.

**Is there a free tier or trial?** Not in the private alpha. Pricing is set on
a call (https://cal.com/boomage-engineering/30min); these rates are the indicative seed.

**Is there a hosted version?** In progress. The alpha runs on your own
Linux + KVM host with our help.

## Limits

- Sandbox sizes: 1, 2, 4, 8, 16, 32 vCPU; 2, 4, 8, 16, 32, 64 GiB memory.
- Scale is up only on a running VM; going smaller means a new machine from a
  commit at the smaller size.
- No free tier and no self-serve signup in the private alpha; join it on a
  call (https://cal.com/boomage-engineering/30min).