Budget & Quota
Every license includes a monthly usage quota for AI requests. This gives everyone in the workspace fair access to all available AI models, from the economical ones up to the most powerful.
The vast majority of users never reach their quota. In practice, the limit only becomes noticeable with particularly intensive use of the most expensive premium models.
Quota per license
Section titled “Quota per license”Every license has its own monthly quota. The higher the license tier, the more room you have:
| License | Quota (multiplier) |
|---|---|
| Pro | 1x (base) |
| Business | 3x |
| Max | 12x |
We deliberately set the quotas generously. Specifically, the limit per license is chosen so that the token costs of the AI providers do not exceed the license price. Most users rarely use up their quota. This is how the model works economically for both sides: you have enough room for your daily work, and we can sustainably offer the service at a fair price.
The quota is reset on the 1st of each month.
Why no hard request count? A number like “X requests per month” would be misleading. A single long analysis with a top-tier model can consume as much quota as several hundred short everyday questions with an economical model. Instead of a number that does not hold up in practice, you can see the real usage of your workspace at any time in the Usage analytics.
Model choice influences usage
Section titled “Model choice influences usage”Not every AI request consumes the same amount of quota. In the model selector and in the model settings, you see a cost category next to each model:
| Category | Meaning |
|---|---|
| € | Economical, good for simple tasks and routine requests |
| €€ | Balanced, good balance of quality and usage |
| €€€ | High quality, significantly higher quota usage per request |
| €€€€ | Top-tier, very high quota usage per request |
The principle: Premium models (€€€ and €€€€) deliver the best results, but consume significantly more quota per request than economical models. Anyone who regularly works with the most expensive models will exhaust their quota correspondingly faster.
Tip: For everyday tasks, a € or €€ model is often enough. Reach for €€€ and €€€€ models when the highest quality really matters, for example for complex analyses, demanding coding tasks or particularly sensitive texts.
Budget sharing within the team
Section titled “Budget sharing within the team”Quotas are shared within the workspace. This means:
- Users who consume less automatically free up room for colleagues with higher demand
- A single user can use up to 200% of their own license quota, as long as the overall budget of the workspace allows it
- This ensures that unused quota does not expire, but benefits the team
This personal limit is the boundary of what one individual can draw from their colleagues’ quotas. Anything consumed beyond it is paid from the workspace credit. They keep working without restriction as long as credit is available; if there is none, only the budget models remain. In the usage analytics such a user is flagged red in the Quota column, with a note that their usage is running on credit.
Checking where you stand
Section titled “Checking where you stand”You can look up your own status at any time, no administrator rights needed: under Settings → Personal settings, the Usage card shows where you currently stand and the date your quota renews.
The percentage is your spend measured against your own quota — the same figure administrators see about you in the usage analytics.
It can go past 100 percent, and that is not an error: beyond your quota you keep working normally, and the excess then runs on the workspace (see Budget sharing within the team). Only once nothing is left there either are you down to the cheap models, and the card says exactly that when it happens.
As you approach the point where things really do stop, a note also appears in the left sidebar. As long as you are comfortably below it, the sidebar stays quiet and you simply keep working.
Cost cap per agent
Section titled “Cost cap per agent”With the Agent platform, tasks can also run autonomously (via schedule or webhook). To ensure that an agent never unnoticed consumes more quota than intended, every agent has its own monthly cost cap as a safety net. When the cap is reached, the agent automatically pauses and the owner is notified by email.
Only autonomous usage counts towards it. What a person writes in the chat with the agent counts towards that person’s quota and does not brake the agent.
The preset is €10/month, which is comfortable for most use cases. You can adjust the value per agent at any time: upwards for more resource-hungry agents, downwards for an even tighter safeguard.
Configurable per agent:
- Monthly cost cap: Owner and administrators can change the value
- Early warning at utilization: Email notification at, for example, 80 % of autonomous usage
- Allowed models: for example only allow economical models, no premium models for high-volume crons
- Auto-pause after 3 consecutive failures: protects against runaway configurations
Where usage is billed depends on what happened, not on who owns the agent:
- Autonomous runs (schedule, webhook, “Run now”): paid out of the workspace’s credit
- What a person types in the chat: counts towards that person’s quota, in whichever agent they typed it
Once the credit is used up, no autonomous run starts any more. A run already under way is finished, and you keep working in the chat as usual.
Administrators see an overview of all agents with autonomous usage, cap and status under Settings → Usage analytics → Agents. They can override the cap per agent specifically.
Controlling usage per user
Section titled “Controlling usage per user”As an admin, you control usage in the workspace specifically where it arises. Two levers are available to you:
- License choice per person in the user management. Pro, Business or Max determine the base quota and which features (agents, integrations, API) are unlocked.
- Agent cost cap per agent in the Usage analytics. You can set the monthly cap per agent as an override and additionally restrict the allowed models per agent, for example only economical models for high-volume crons.
Which brake kicks in when?
Section titled “Which brake kicks in when?”In the workspace there are several independent brakes that safeguard usage. They act separately from each other, which is the most common point at which the question arises: “My budget has been raised, why is my agent still not running?”. The following table makes the mapping clear:
| Brake | When it kicks in | What happens | What you can do |
|---|---|---|---|
| Workspace quota exhausted | When the monthly workspace pool of all licenses is used up | All users keep chatting with the economical models, premium models are disabled until the next reset | Top up credit, upgrade license or wait until the 1st of the next month |
| Credit used up | When the workspace’s credit is spent | Autonomous runs no longer start. You keep working in the chat as usual, as long as the license quotas last | Top up credit |
| Agent cost cap reached | When the monthly autonomous usage of an individual agent reaches its own cap (default €10) | The agent runs no further autonomous runs, owner and admin are notified by email. Other agents keep running, and you can still chat with it | Increase the cap in the Models & cost tab, or wait until the end of the month. Admins can override the value in the usage analytics |
| Auto-pause after three consecutive failures | When an autonomously triggered agent ends with an error three times in a row | The agent pauses, further cron and webhook runs are not executed. Chat runs are not affected | Check the cause in the run history, adjust the configuration and reactivate the agent in the General tab |
| Daily limit during the trial | During the 7-day trial, when the daily quota is reached | Hard limit: new AI requests pause until the daily reset | Wait until the next day or purchase a license |
Practical consequence: Increasing your personal share of the workspace pool does not help if an individual agent has reached its own cost cap. These are separate levers. If an agent unexpectedly pauses, first look at the agent cost cap in the Models & cost tab and at the status in the Usage analytics.
What happens when the quota is exhausted?
Section titled “What happens when the quota is exhausted?”Even when the quota is used up, you remain able to work:
- You keep chatting, but only with the economical models. These easily cover the vast majority of everyday questions.
- The expensive premium models remain disabled until the next quota reset. In the model selector they appear greyed out, with a note indicating when they will be available again.
- The platform remains fully usable: knowledge management, settings and reading existing chats are possible without restriction.
- A status notice with the reset date appears in the left sidebar.
If you want to use the premium models again immediately, you have two options:
- Top up credit to continue working straight away
- Upgrade to a higher license for permanently more quota (for example from Pro to Business or Max)
Otherwise, the premium models are automatically unlocked again on the 1st of the month, when the quota resets.
Note on autonomous agents: Cron or webhook agents only keep running in this state if they are configured for one of the economical models anyway. If the agent uses a premium model, it pauses until the next reset or until you adjust the model. More on this under Cost cap per agent.
Credit
Section titled “Credit”Credit is a workspace account that administrators can top up at any time under Billing. Two things are paid out of it:
- The entire usage of autonomous agent runs (schedule, webhook, “Run now”)
- What individual users consume beyond their personal limit (see Budget sharing within the team)
- Everything the workspace consumes beyond the sum of all license quotas
In the quota card of the usage analytics, the line below the credit balance names what it went towards this month, with a separate amount per source.
Unused credit does not expire, it remains available in the next month.
How topping up works
Section titled “How topping up works”Under Billing you choose the number of packages and are taken to the payment page. Once the payment goes through you land back in the application, the credit is available immediately, and the invoice is sent automatically to your billing address. If you abandon the payment, nothing happens: no order and no invoice are created.
When paying by card, your bank may ask you to confirm the purchase, usually through your banking app or by text message. Whether it does is the bank’s decision.
If you pay by direct debit or on account, it works the other way round. You receive the invoice with the order, including a payment term, and can settle it at any time via the link in your top-up history. We enable these payment methods on request, so talk to us if you would like to use them.
Once the credit is empty, the autonomous agents pause while work continues in the chat, as long as the license quotas last.
Credit is particularly suitable for months with increased demand, for example during projects, quarterly closings or when new employees are intensively getting to know the platform.
Topping up automatically
Section titled “Topping up automatically”So that agents and the team can keep working, topping up can be automated. In the credit card under Billing, you set three values:
- Threshold: the remaining balance at which credit is bought
- Amount: how much is bought then, in steps of €10 and at least as much as the threshold
- Monthly limit: how much is bought automatically per month at most (optional)
The amount has to reach the threshold on purpose: otherwise the balance would still be below it after the purchase and credit would be bought again straight away. All amounts are net; the charge is the gross amount including VAT.
The charge goes to the same payment method your subscription uses; no new payment details are needed. Switching it on asks you once to authorise those charges, and you can switch it off again in the same place at any time. The feature is off by default, and only someone who may manage the billing can change it.
Card or direct debit is required. If you pay with Amazon Pay or Klarna the switch stays locked, because those methods are not suited to irregular amounts charged without your involvement. Changing the payment method is all it takes.
The balance is checked every five minutes. Once it falls below the threshold, credit is bought exactly once, not repeatedly. When the monthly limit is reached, nothing more is bought; the usual warning that credit is running low takes over instead.
When a charge fails
Section titled “When a charge fails”With a card, neither credit nor an invoice is created. All administrators receive an email, and automatic top-up pauses until a purchase by hand has succeeded. If the card is declined permanently, because it has been blocked for instance, the automation switches itself off entirely: a card that is charged again and again eventually gets blocked by the bank.
With direct debit, credit and invoice are there straight away, because the outcome of the charge is only known days later. If the debit is returned later on, the credit stays, our support gets in touch, and the automation pauses.
The credit card always shows when credit was last bought automatically and how much has been added automatically this month.