> ## Documentation Index
> Fetch the complete documentation index at: https://nemonist.com/docs/llms.txt
> Use this file to discover all available pages before exploring further.

# Credits and billing

> Provider prices on every token. Start free; a fee is taken only when you buy credits.

Every round is charged at the provider's list price for its input, output and cached tokens. There is no markup on tokens. Learning is included: memory rounds are charged like any other round. A memory round served by your own provider key is billed by your provider, not nemonist, and counts toward your key's spending limit. On **Logs**, each reply also shows about how many of its input tokens nemonist added for memory (*by memory*), since you pay for those too.

## Free

Free runs on your own keys (see [Bring your own key](https://nemonist.com/docs/concepts/byok.md)): your own Router key for any model, or provider keys with shared capacity off for their models, when your account with that provider has zero data retention (Anthropic and OpenAI retain prompts by default, so a standard pay-as-you-go account does not qualify). Otherwise it may use free models (ids ending in `:free`) that keep no data, when any are offered, at 50 requests a day; there may be none. No credits are given or needed.

## Buying credits

On the **Credits** page, buy from $5 to $500 at a time with a card (paid through Stripe). The plan's fee and the card processing fee are both shown before you pay. Unused credits stay on your account.

* **Automatic top-up** buys credits when your balance falls below a threshold, charging your saved cards in order (up to three). Optional and off by default.
* **Promo codes** add credits to the account once.
* **Refunds** return the fees with the credits they take back. To delete an account with a balance, contact us for a refund first.

The **Credits** page lists every purchase, automatic top-up, promo code, refund and adjustment.

## Plans

| | Free | Standard | Teams | Enterprise |
| - | - | - | - | - |
| Fee on credits bought | None | 5.5% | 8% | Custom |
| Workspaces | 5 | 50 | 1,000 | Custom |
| Requests a minute | 600 | 600 | 3,000 | Custom |

Teams adds advanced notifications, [Workload identity](https://nemonist.com/docs/enterprise/workload-identity.md) and opt-in [team memory](https://nemonist.com/docs/concepts/memory.md#team-memory). Enterprise adds invoicing, a credit line, [SSO](https://nemonist.com/docs/enterprise/sso.md) and [SCIM](https://nemonist.com/docs/enterprise/scim.md). The full comparison is on the [pricing page](https://nemonist.com/pricing).

To move to Teams, buy credits on Standard, then the account owner chooses **Move to Teams** under **Credits**. The new fee applies to credits bought after the move.

## Managed compute

Your agent can run jobs on nemonist's cluster with the `jobs` tool. Jobs need Pro or credits; Free has no managed compute.

| Queue | Each node | Price per node-hour |
| - | - | - |
| `cpu` | 2 CPUs, 7.6 GB | \$0.2040 |
| `gpu` | 1 NVIDIA L4 GPU (A10G when L4 is short), 2 CPUs, 15 GB | \$0.9658 |
| `cpu-large` | 192 CPUs, 747 GB | \$13.3540 |
| `gpu-large` | 1 NVIDIA L4 GPU (A10G when L4 is short), 4 CPUs, 30 GB | \$1.1731 |

These are the cloud list price plus 20%. A job is billed by the second: its cost is its run time x its nodes x the queue's price. On the project's **Jobs** page, each ended job shows its run time to the second and its cost.

When a job is submitted, its time limit x nodes x price is held from your balance and limits. When it ends, it settles to what it ran, and the rest is released. A cancelled job is charged only for the time it ran.

Pro includes $5 of managed compute each month. Past that, jobs draw from the account's monthly **compute budget**, which starts at $0. The account's owner or an admin sets it under **Credits**. A job that does not fit is refused before it runs. The refusal names every limit it would exceed (the key's spending limit, the included compute, the compute budget) and where to raise each one.

### Job secrets

A job that needs a token (a model hub, a dataset, a private repository) gets it from the project's **Job secrets**, on its **Jobs** page. Add a name, such as `HF_TOKEN`, and its value: every job submitted after that gets it as an environment variable of that name. Values are encrypted for this project and name, and no page, API or agent tool shows them again; you see only names, when each was set and by whom. Each value is replaced with `***` in a job's copied output. A project holds up to 100 secrets. Deleting the project deletes its secrets; when you merge a project into another, its secrets are deleted too (each is encrypted for its own project), so add them again on the project you keep. Never put a token in a job script: ask your agent to use a secret instead.

## Spending caps

* A key's own [spending limit](https://nemonist.com/docs/concepts/api-keys.md#limit-a-key).
* A [workspace's budgets](https://nemonist.com/docs/concepts/workspaces.md): daily, weekly, monthly or lifetime.
* [Guardrails](https://nemonist.com/docs/concepts/guardrails.md) with a budget per key.

A request over a cap, or with no credits left, is refused with `402`. Set alerts for a low balance and budgets under **Notifications**.


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.