Usage limits in Langdock
Every Langdock plan comes with a fixed allowance for AI requests. It applies in two windows at once: five hours from your first message, and a week that starts over every Monday. Once either one is full, GPT-5.6 Luna answers until exactly that window starts over.
By Dennis Cutraro, Co-Founder and Managing Partner · As of:
Why Langdock has limits
Every request to an AI costs compute. It is billed in tokens, and every model has its own price per token.
You, on the other hand, pay Langdock a fixed price per seat. That only works because every seat comes with a limited amount of usage.
This keeps costs predictable, and a few heavy users cannot quietly use up the allowance.
How Langdock measures usage: tokens
Langdock does not count your messages, it counts tokens: what is in the prompt plus what is in the answer, times the price of the model.
A token is a piece of a word, often a single syllable. A short question to a small model therefore costs next to nothing.
A long chat with an attached file and a strong model, on the other hand, pulls noticeably on the allowance, and tools like web search come on top. Langdock shows how much each model uses right in the model picker.
The two time windows for Langdock usage
Both windows run at the same time, each with its own limit. The session hangs on your first message: five hours from then, breaks included. The week is a fixed calendar window and starts over every Monday.
Session
The window is tied to your first message, not to the clock. After the reset, the next message starts a new one.
Week
Today is
Today is Wed. The next reset is Monday, in 5 days. Until then your weekly limit still applies.
Fixed calendar window, independent of activity.
Langdock token limit reached: what happens
You can keep working. Langdock sends your messages to GPT-5.6 Luna, an economical model, until the full window resets. You can check how much you have used at any time.
Premium models
free model choice
GPT-5.6 Luna
automatic fallback
At 100 percent GPT-5.6 Luna takes over. With the reset the bar is empty again.
Check your own usage
- Indicator next to the model picker
- Settings › Account › Usage
Cost control: extra usage in Langdock
Without extra usage, the premium models stop at the limit. Work continues on GPT-5.6 Luna.
With extra usage, the premium models keep running. Past the limit you pay by consumption at API rates, as a separate line on the monthly invoice next to the seats. What each model costs per million tokens is listed in Langdock's model overview.
So the invoice does not get out of hand, admins set caps on three levels that can be combined. The amounts in the planner are an example.
Plan the budget
The cap for the entire workspace. Once reached, extra usage ends for everyone, regardless of plan and user limits.
Applies to every user on the plan who has no individual user limit.
An individual limit that admins assign to specific users, such as heavy users. For them it applies instead of the plan limit.
For example AI champions and heavy users.
12 users in the workspace
- 0 €
- 20 €
- 20 €
- 0 €
- 20 €
- 20 €
- 0 €
- 20 €
- 20 €
- 0 €
- 20 €
- 20 €
160 € / 150 €
Too much. The individual limits add up to more than the workspace limit. Once it is reached, extra usage ends for everyone, including users who have not used up their own budget.
Example values. Amounts are per month, the limits reset with each billing period.
Frequently asked questions
Sources
- docs.langdock.com: Extra Usage
- docs.langdock.com: Usage and spend limits
- docs.langdock.com: Usage Limits (Fair Usage)
- langdock.com/pricing: Preise (Stand 08/2026)
- langdock.com/models: Modelle und Preise pro Million Tokens
Author
Dennis Cutraro is Co-Founder and Managing Partner at Unfuture. He talks to customers about Langdock rollouts every day and is in ongoing contact with the Langdock team.
