# Plans and limits

> What each emit cloud plan costs, what it includes, and the limits that apply today.

Every cluster is on one plan. All plans use the same API, SDKs and dashboard. Prices are in US dollars, and you're invoiced monthly.

## Plans

| Plan | Price | Jobs in flight | Claims a minute |
| --- | --- | --- | --- |
| Free | $0, for 1,000 jobs a month | 10,000 | 600 |
| Serverless | $5 a month per cluster, plus $3 per million jobs | no limit (fair use) | 60,000 |
| Dedicated xs | $20 per node a month, no per-job charge | no plan limit | no plan limit |
| Dedicated s | $40 per node a month, no per-job charge | no plan limit | no plan limit |

**Jobs in flight** are jobs created but not finished: pending or running. At the limit, new jobs wait until workers bring the count down: emit answers `429` with a time to wait, and the SDKs try again for you, for up to a minute unless you set longer. See [When emit is busy](/docs/concepts/jobs#when-emit-is-busy).

### Free

For trying emit and for small projects. It needs no card. When a Free cluster has used its 1,000 jobs for the month, it's paused until the month turns, or until you upgrade it to Serverless. While it's paused, it takes almost no new jobs and hands almost none to workers. The jobs already there are kept.

### Serverless

Your cluster runs on emit's shared machines, and you pay for the jobs you create. The $5 fee is charged for the rest of your billing month when you create the cluster, then on each monthly billing date. It isn't refunded if you delete the cluster. Jobs are invoiced monthly. A Serverless cluster needs a payment method on file.

Serverless has no limit on queued jobs. Fair use applies to very large backlogs: past it, new jobs are held back until your workers catch up. The SDKs wait for room for you, for a minute by default. Contact us if you need more. Each cluster includes 5 GB of archive storage; beyond that, it's $0.03 per GB-month.

For example, 20,000 jobs in a month cost $5 + $0.06 = $5.06. At 5 million jobs a month, Serverless costs the same as one Dedicated xs node: $20.

### Dedicated

Your cluster runs on its own nodes, with no per-job charge. At a node's memory or disk limit, new jobs wait.

| Size | Price | Throughput (estimate) | Included each month, per node |
| --- | --- | --- | --- |
| xs | $20 per node a month | ~500 jobs/s | 5 GB archive, 10 GB egress |
| s | $40 per node a month | ~1,000 jobs/s | 10 GB archive, 20 GB egress |

Throughput figures are estimates for jobs created in batches. Creating one job a request gets about a third.

Run one node, or three across three availability zones so the cluster survives losing one. Nodes are billed for the time they run, at the monthly price spread over 730 hours. Beyond what's included, archive storage is $0.03 per GB-month and egress $0.10 per GB. Larger sizes are coming.

A Dedicated cluster starts once your organisation has a payment method. Without one, setup fails after 3 days and the cluster has to be created again.

## What counts as a job

On Free and Serverless, every job created counts: a job you create, and in a [workflow](/docs/concepts/workflows), each run, each step, each `wf.do` and each signal. Sleeping, waiting and picking a run back up are free. Retries of a job don't count as new jobs.

## Regions

Clusters run in **US East (N. Virginia)**. Your data stays in the region you choose.

## Limits

These apply to every plan.

| Limit | Value |
| --- | --- |
| A payload, a result, a signal's data | 4 MiB of JSON each |
| A job name, an idempotency key | 256 bytes each |
| An event | 4 KiB; 1,000 a job unless the space allows up to 3,000 |
| A log line or a progress message | 4 KiB |
| An error message | 4 KiB; longer is cut |
| Attempts of a job | 1 to 100 |
| The longest wait between retries | 1 hour |
| A delay, or a scheduled time | up to a year ahead |
| A job nobody claims | fails after 7 days, unless the space says otherwise |
| An idempotency key | holds while its job is unfinished, and 10 minutes after it finishes |
| Jobs claimed in one call | 1,000 |
| A claim's wait for work | 60 seconds |
| A callback | retried for about a day |

Workflows have [limits of their own](/docs/concepts/workflows#limits).