One credit, one question.
Every question costs one credit, so there are no tokens to work out and no surprise bill. See what the company asked, on which models, and by whom. Set a daily allowance per person and a monthly one for the company. Meetings have their own separate allowance, with fifty hours included.
A measure is defined once - the entity, the aggregation, the field and the currency-safe field - and every widget, report and board reads that definition rather than a column somebody picked.
What it costs you
Costs anyone can understand.
Everything is counted in questions, so anyone in the company can look at the number and know what it means.
Ask the assistant something and that is one credit. No tokens or word counts to work out.
Transcription, the write-up and the coaching all run on real AI and none of it touches your credits. Fifty hours come with the plan.
Each person has their own allowance, so one heavy week from one person does not eat the whole team's month.
Point a department at a model running on your own machines and its questions are free.
Watching it
What the company asked, and on what.
Six numbers across the top, three charts under them, and every model, person and question underneath. Choose today, the last seven days, the last thirty, or everything since you started, and the whole screen updates to match.
- GPT-5.448%
- Claude Opus 4.832%
- Gemini 3.5 Flash13%
- Claude Sonnet 4.67%
The six numbers
Every number shows which way it is moving.
Each figure is shown with its change against the previous period, so you can see at a glance whether it is rising or falling.
How many questions the company asked, in the unit you are billed in. Nothing to convert.
How long an answer took, on average. Moving a department to a faster model shows up here within a day.
The total count across everyone, shown next to credits so you can spot when one climbs faster than the other.
The raw usage underneath, for anyone who wants the detail. You are never billed on this number.
How many different models the company actually used. Useful the week after you change which models answer.
The share of requests that failed. A provider having a bad afternoon is visible before anyone reports it.
The charts
Three charts answer the big questions.
Is it fast, where is it being used, and what is the expensive part. One chart each, drawn from the same period you picked at the top.
Not just the average. The typical answer, the slow one in twenty and the slowest in a hundred, plotted together, so a few slow answers cannot hide behind a good average.
One bar per day, split by department. You can see a team start using the assistant, and you can see one stop.
The share of credits each model used over the month, so you can see whether the expensive model is earning its place.
Fifty hours of meetings, outside your credits.
Meetings
Meeting AI has its own allowance.
Transcription, the write-up and the coaching afterwards all run on real AI, and none of it touches the credits your people ask questions with. Fifty hours come with the plan, so a heavy month of meetings never cuts into the assistant allowance.
Recorded and transcribed meetings, bundled with the plan rather than drawn from anybody's allowance.
A summary, the decisions and the follow-ups, produced by a real model reading the transcript.
Talk ratio, questions asked and the habits worth changing, scored after the call.
A heavy month of meetings leaves the assistant allowance exactly where it was.
Three limits, and you choose all of them.
A daily allowance per person, a monthly one for the company, and a rate per minute.
- Cap what any one person can use in a day
- Cap what the whole company can use in a month
- Cap how many questions can be asked per minute, so a runaway automation cannot drain credits
- Set any of them to unlimited
- See the same meter your people see, so nobody is surprised
Everyone has their own allowance.
Your meter and the limit that applies to you are the same number, so you always know where you stand.
- See your own use as you go, in credits
- Keep an allowance that is yours alone, not shared from a pool
- Let a manager see what their department is using
- Compare what each model actually cost you across a month
- Watch a department move to your own hardware and its cost drop to zero
Spending less
It shows you where the credits go.
NearSync looks at your own usage and points out the obvious savings: a department using a heavy model for simple questions, a model that costs several times more than an alternative for the same work, or a team that could run on your own hardware for free. Make the change on the Setup screen, and the next chart shows whether it worked.
Named, with the model it uses now and the kind of questions it asks.
The cost per question, shown next to the model you would switch to.
The saving in credits, over the period you are looking at.
Access
Who can do what.
Everyone can see their own meter, and managers can see their department. What the whole company uses, and the limits themselves, belong to Admin.
Watch your own credit meter as you use it.
Ask questions, knowing exactly how much of your allowance is left.
See what your department is using, and on which models.
Set the daily, monthly and per-minute limits.
Admin is the highest role and is not constrained by the permission grid, so it is worth keeping to the smallest group that genuinely needs it.