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

# Plans & limits

> What each tier allows, and what happens at the boundary.

Plans are flat subscriptions, not credits. Model spend goes to **your own**
provider connection, so there is nothing for us to resell - see
[Models & providers](/docs/concepts/models).

|                    | Free | Pro    | Business  | Enterprise |
| ------------------ | ---- | ------ | --------- | ---------- |
| Price / month      | \$0  | \$99   | \$499     | Custom     |
| Members            | 2    | 10     | 50        | Unlimited  |
| Agent runs / month | 500  | 10,000 | 100,000   | Unlimited  |
| Knowledge bases    | 1    | 25     | Unlimited | Unlimited  |
| Enabled schedules  | 1    | 25     | Unlimited | Unlimited  |

Current pricing always lives at [cruq.ai/pricing](https://cruq.ai/pricing), and
billing is managed under **Workspace → Billing**.

## What counts as a run

Every agent execution: the playground, `/v1/agents/{id}/run`, a schedule firing,
an inbound channel message, and **each cell of an eval experiment**. A 5-model
x 20-case comparison is 100 runs.

## At the boundary

Limits are checked **before** work starts, not after:

* A workspace over its run quota is refused before a run row is written, so a
  blocked run does not consume the quota that blocked it.
* An eval experiment is quota-checked before any of its cells are created -
  either the whole grid is created or none of it is.
* Seats are checked when an invite is created **and** again when it is accepted,
  since a workspace can fill up in between.

Nothing silently overages. You get an explanation naming the limit you hit.

## Eval budgets

Independently of your plan, an experiment can carry a `budget_usd`. When spend
on that experiment passes it, the remaining cells are skipped and the reason is
recorded. This bounds a single comparison; the plan quota bounds the month.