AI work costs real money, and nobody likes surprises. This update makes the cost of every run visible while it happens, shows your remaining budget at a glance, and makes sure a run cannot spend more than you have.
Cost on every turn
After each AI turn you now see a single line with everything that matters: how long it took, tokens per second, tokens in and out, and the cost. When a run finishes there is one more line with the total for the whole run, with the same details added up.
The cost appears as soon as the provider reports it. While it is still on its way, the line shows a small spinner instead of a "loading" word, so you can tell waiting from a missing number.
We also removed our own token estimates. Different providers count tokens differently, and an estimate that disagrees with your bill is worse than none. The numbers you see now come from the provider.
The fuel gauge
On the small rail beside the editor there is a new gauge, a tank of liquid in the style of the rest of the interface. It shows two things:
- your monthly allowance from your plan, which refills every month, and
- your credits, which you can buy at any time and which never expire.
Above it a badge shows your plan. Below it the numbers are written out, for example 2.48 / 3.00 with the limit in grey. The gauge replaces the old budget panel above the chat input, which is gone.
Runs stop when the budget is gone
A long task, with several agents looping and many tool calls, can cost a lot more than a single question. So the limit is enforced during the run rather than only afterwards. Each account has its own spending limit at the AI provider, and when it is reached the run stops with a clear message: AI budget used up. Top up credits or upgrade your plan to continue.
The step that was interrupted is not shown as a red failure, because it did not fail; it just ran out of budget.
Costs that arrive late
Sometimes the exact cost of a request only reaches us hours after it ran. We now take those late costs into account. If one arrives after you have already used your balance, your balance can go slightly negative, shown as an amount owed. The next time you add credits or your monthly allowance refills, the amount is taken from it first. Nothing is lost and nobody is charged twice.
Every model, for everyone
The model selector now lists all available models, including for Free and demo accounts. Choosing a paid model on an account with no budget gives the clear budget message above instead of a hidden restriction. Demo accounts have a budget of zero and cannot spend anything.
For administrators
There is a new admin dashboard page that gathers the plan settings and the AI model settings in one place. Edits to the platform's default pipeline now reach every user on their next message, with no reload needed. We also fixed the admin hosting page, which could slow the whole server down while it loaded; it now loads quickly in the background.
You can see your plan and what is left in the gauge right now. See the new plans for what each one includes.