Plans, quotas and billing

The Billing page at app.xicbot.de/abrechnung shows which plan your account is on, how much of the monthly quota has been used and which line items an invoice is made up of. This page covers the three plans Start, Pro and Custom with their limits and features, the monthly quota with its warning threshold and throttling, and every field of the view. Prices are listed separately under Pricing.

Plans and their limits

The limits appear in the Billing view under Plan limits, with the same names as in this table. If an account carries a plan the service does not recognise, the Start limits apply, and the view notes this below the usage bar.

• Seats count the people in the account plus any open invitations. An invitation takes up its seat before it has been accepted, see Team and roles.
• Assistants and Knowledge sources count every entry in the account, whatever its status.
• Documents and Sections only come into being when a source is read in; how many depends on the size of the source. Details are under Knowledge.
• Units per month is the monthly quota, explained under Monthly quota.
• Requests per minute caps calls from the chat window per visitor, explained under Requests per minute.

When a count limit applies

The service checks the limits for seats, assistants, knowledge sources, documents and sections before anything is created. A limit counts as reached as soon as the existing number equals it: on Start, the first assistant already reaches the limit of 1, so a second one cannot be created.

• Documents and sections are only created while a source is read in. Before a source is created and before it is read in again, the check therefore asks whether the limit has already been reached. After reading in, the service checks the new count a second time before it takes effect: if an intake or refresh would increase the documents or sections and take them past the limit, it is not applied, and the previous state of the source is kept. This also applies to refreshes in the background.
• Entries above a limit are kept, for example after a move to a smaller plan. Only creating new ones is blocked.
• To make room, delete knowledge sources you no longer need (role Edit) or assistants (role Administration), or withdraw an open invitation (role Administration). Deleted entries no longer count, and removing a knowledge source also removes its documents and sections.

Through the API, a reached limit returns status 402 with plan_limit_reached, plus limit naming the limit (such as assistants or seats), allowed, used, plan and plan_name. If the count cannot be taken, the response is 503 with plan_limit_unmeasured. POST /api/v1/integration/knowledge/sources, which creates a source with an API key, checks the same limits for knowledge sources, documents and sections.

Plan features

Besides count limits, features unlock entire areas. Which features an account has depends solely on its plan; the role additionally decides who may use the area.

Answers in several languages are not tied to any feature and are available on every plan, Start included. There is no limit for them and no separate line item per language: a reply counts against the quota with its units, whatever language it is written in.

Monthly quota

The quota under Units per month counts units (tokens): those sent to the language model and those it returns. Counting is per account and calendar month, across all assistants together. A new count starts on the first of each month, and the previous month stays listed under History.

• Replies count in the chat window on your website, in the Test dialog, in Acceptance runs and through the API.
• A Setup run counts with its units and as one reply.
• A question under Check an answer on the Knowledge page counts with its units and as one reply whenever a model uses units to answer it. See Knowledge.
• Messages in a conversation that a person has taken over do not count, because no model is asked there.
• An answer from an active question and answer pair does not count: it goes out word for word without a model being asked. Once the quota is exhausted, however, such a question also gets the fallback answer. See Knowledge.
• If the stored quota is 0, it is treated as exhausted from the start.

Before a reply is generated, the service reserves 2,500 units, or 12,000 for a setup run. A reservation is only made while usage is below the quota. Once the reply is done, the actual usage replaces the reservation; if no reply comes about, the reservation is released again. If a run is cut off before booking, the reservation stays in that month's usage. While a reply is running, its reservation is part of the usage shown. Because booking happens after the reply, usage can exceed the quota by the replies that ran last; the excess appears under Line items.

Warning threshold and throttling

The Warning threshold sits at 80 % of the quota, rounded up to whole units: 400,000 on Start and 2,400,000 on Pro. From there on, the assistant keeps answering in throttled mode. Throttling already applies to the reply whose reservation crosses the threshold, and it ends when the month changes or when a larger quota brings usage back below the threshold.

Crossing the Warning threshold creates a notice for the people who run XICBOT; as long as usage stays above it, no further notice follows. If a single booking jumps past both the threshold and the quota, only the notice about the exhausted quota is created. These notices go to the operator, not to the users of your account. You can see where you stand under Billing in the notices at the top. To get a message yourself when the quota is reached, use the Budget exhausted webhook; see When the quota is exhausted.

When the quota is exhausted

The quota is exhausted as soon as usage reaches it. From then on the service asks no model and books nothing. What happens instead depends on the channel:

The API response looks like this; month is the current billing month, budget the quota and used the usage so far:

{
  "error": "quota_exhausted",
  "scope": "month",
  "month": "2026-09",
  "budget": 500000,
  "used": 502140,
  "retry_after": 86400
}

If you have set up a webhook under Interfaces for the Budget exhausted event (budget.exhausted), you receive a message with month, budget and used at the booking that makes usage reach the quota. No further message follows while usage stays above it. Webhooks are part of Interfaces and are therefore only available on the Custom plan. See API keys and webhooks.

Replies resume as soon as the next month begins or the quota is raised above the current usage.

Requests per minute

Requests per minute caps how often a single visitor can call the chat window within one minute, for example to load its settings or the history. Calls are counted per visitor and additionally per conversation, and either counter can refuse. Sending questions has its own limits. All values and the response when a limit is exceeded are described under Widget delivery. API keys carry their own per-minute value per key, see API limits.

The Billing view

Billing sits in the sidebar in the Administration group, together with Team and Account. Every role can open it once the email address has been confirmed; until then it reads “This area stays locked until the email address is confirmed.” The page is headed Usage and billing with the sentence “The usage of the current month, the limits of your plan and the line items an invoice is built from.” It is for viewing only; nothing can be changed here. Below Payment follows the Change plan section: “The plan cannot be changed in the app; we switch it for your account. The change is requested by whoever signs for the account, by email to the address below. Please state the account name and the plan you want.” Below it are the address mail@xicbot.de as an email link with the subject “Plan change” and the account name, and the link What a change affects to this page. The section is shown even if the overview could not be loaded.

Notices at the top

“Replies” here means replies from the model. What visitors and the API receive instead is described under When the quota is exhausted.

Usage

• Below that, Plan shows your plan name, usage and quota (such as 412,300 / 500,000) and a bar with the share in whole percent, truncated rather than rounded and capped at 100. The bar changes colour from the warning threshold on and turns red once the quota is exhausted.
• With a quota above 0, Warning threshold below the bar gives the number of units from which throttling applies. If your account has a plan the service does not recognise, it shows “This value could not be determined and is therefore not a result.” with the stored key in brackets.
• The Warning threshold card lists the state for each measure, for example “412,300 of 500,000 · 82% · Warning threshold: 400,000”. Here, as on the bar and in the notice at the top, the share is truncated rather than rounded. The Replies billed row reads Not measured with the reason “No monthly quota of answers agreed. Usage is counted, not measured against a limit.”, since there is no limit for replies.
• If the stored quota is 0, the Units billed row reads Not measured with the reason “No quota stored. There is nothing to warn about before the limit; the limit has already been reached.”

Plan limits

The Plan limits table has the columns Name and Value and lists the eight limits from Plans and their limits. Numbers are shown with thousands separators, Conversation export as Yes or No. For Units per month and Requests per minute, the table shows the values stored for your account on every plan.

Line items

Above the table, the Period runs from the first to the last day of the month, followed by the price basis, for example “Period: 01/09/2026 – 30/09/2026 · Amounts net of VAT”. The table has the columns Item, Quantity and Amount.

• Total adds up the amounts. If any item has no amount, it reads Not measured followed by “At least one agreed price is missing. The total is therefore not formed rather than invented.” This applies to the Custom plan and to every month with units above the quota.
• Under Payment it says “No payment provider is connected. Billing happens outside this interface.” Nothing is paid and no invoice is triggered inside the app.

History

History shows up to 24 months, newest first and including the current one, with the columns Month, Units billed and Replies billed. Without any usage it reads “No usage recorded yet.”

When a limit is not enough

The plan cannot be changed inside the app; it is set for your account. To request a change, write to mail@xicbot.de; the email link for this is under Billing in the Change plan section. After the change, and as described in the note on stored values, the new limits apply to the next item you create and the new quota to the next reply. Usage in the current month carries on; an exhausted quota allows replies again once the new quota is above the usage.

API

Besides 402 for a missing feature or a reached limit, the plan check has two more answers. If the request belongs to no account, it returns 403 with no_tenant. If an endpoint checks a feature or limit that no plan knows, it returns 500 with plan_gate_misconfigured, plus feature, limit or kind naming it. That is an error on the XICBOT side. In both cases the request is not carried out; please report the second case to mail@xicbot.de.

Frequent questions