AI Turns & Usage
Hosted AI is metered by turns.
| Plan | AI turns per month |
|---|---|
| Free | 50 |
| Plus | 250 |
| Pro | Unlimited |
What counts as a turn
Section titled “What counts as a turn”A turn is one call to the AI. Most of the time that is exactly one exchange — you send a message, the agent answers, that is one turn. A long question does not cost more than a short one, and attaching an image or a file does not cost extra.
Sometimes one message costs two. If a request needs the stronger model, or the agent has to read data from one of your source cards before it can answer, it makes a second call — and that is a second turn. This is uncommon and it is not something you need to manage.
A multi-step request costs one turn per step. When the agent works through several steps to reach an answer — researching, then reading your workspace, then writing a file — each step is its own call, so a five-step request uses five turns. The step count is shown as it runs, so you can always see what a request is costing while it happens rather than afterwards.
Where to see what is left
Section titled “Where to see what is left”The remaining count appears near the composer whenever your plan is metered. Pro accounts and BYOK plans do not show a meter, because there is nothing to count.
When you run out
Section titled “When you run out”Hosted AI pauses until the next reset. Nothing else stops.
Your workspaces, your cards, your connected sources and their scheduled refreshes all keep working exactly as before. Vlynks keep watching. The only thing that waits is the agent’s ability to answer.
Usage resets at the start of each calendar month, UTC.
Three ways past the limit
Section titled “Three ways past the limit”Wait for the reset — everything else keeps running in the meantime.
Upgrade — takes effect immediately, and you pay the difference. See Plans & Limits.
Switch to BYOK — connect your own API key and stop being metered by Vexlynk entirely. See BYOK.
Using your turns well
Section titled “Using your turns well”Turns are per AI call, so the way to get more out of them is to ask fewer, better questions rather than splitting one thought across five messages.
A question grounded in what is actually in your workspace also tends to need less follow-up, which is the real saving — see Using the AI Agent.
- BYOK — your own key, no Vexlynk metering
- Plans & Limits — everything each plan includes
Was this page helpful?
Still have a question?
Ask it here. Questions shape what gets written next — if something is missing from this page, this is how it gets added.