Try the product on sample data. No account, no key, nothing to install.
You can try Recovea now, without signing up for anything. Open the demo and you land inside the platform, already signed in, with a month of traffic sitting in it.
It is the same software we ship. Not a video, not a click through tour with a highlighted button telling you where to go next. Everything is live, so open whatever you want and take as long as you like.
Recovea is four things, and each one is easier to understand in about ten seconds of clicking than in a paragraph. Here is where to find them.
Meter
Start on the dashboard, then open the requests list and scroll. A month of requests, each one metered as it happened rather than totalled at the end of a billing period.
Open a single request. You get the model it went to, the token counts broken out by class, what it cost, how long it took, and the key and project it came from. The price it was charged against is the published list price, and it is right there on the receipt, so you can check the arithmetic on any row you like.
Then look at the same month split by key, by project and by model. That is the view most people come here wanting, and it is the one your provider dashboard cannot give you, because it does not know which of your things made the call.
Cap
Open the caps and see which are armed and how much room is left on each.
Then set one yourself. Put it somewhere tight, pick the window, and watch what happens to the numbers. A cap in Recovea is not an alert that arrives after the money is gone. When a request would cross the line, it is refused before your provider is ever called, so there is no call, no tokens and nothing on the bill. What comes back is a 402 that names the scope it hit and the window it hit it in, so your code knows precisely which wall it met and can do something sensible about it.
Caps stack. Workspace and single key, on the day, the week, the month, the year, or for the lifetime of the thing. The tightest one wins.
Breaker
Open an agent run and look at the dollar budget it carries.
This is the one that answers the loop running at three in the morning. A run gets an envelope, and when the run reaches it, the next request is refused and that run stays tripped. A sibling run on the very same key keeps its own envelope and carries on working, because the budget belonged to the run, not to the key.
Find the run in the demo that hit its limit and see what that looks like from the inside. It is the difference between a spend control and a power switch.
Prove
Open the ledger.
Every priced request wrote a line, and every line carries a fingerprint of the line before it, so the whole month links into a chain. Refusals get their own row too: the request that did not happen is recorded as not having happened, which is the part most systems simply lose.
Export the receipts and re derive the chain right there in your browser. You are checking our arithmetic without asking us anything, which is the entire reason it is built this way. Keep a copy of the chain head anywhere you like and that copy is your protection against us.
There is more in there than the four
Spend attributed per key, per project and per route. Alerts and where their thresholds sit. Run policy. Notifications, settings, billing, the lot. The demo is the whole platform on a sample workspace, so nothing is stubbed out and nothing is behind a "contact us".
Click anything. Change anything. It costs nobody anything, and closing the tab is the reset button.
Then bring your own numbers
When you have seen it work, the real version is one base URL change and your own provider key. Free is cardless, the meter starts on the first request, and the first receipt usually shows up within about ten minutes.
Open the demo, or read the docs if you would rather see the shape of it first.