Paid plans buy capacity and history, not features, so you can count what you owe before you agree to it: how many people, how many projects, how far back.
One honest exception: the activity log needs a paid plan. On Free the account has a single person, so every line in it would name whoever is reading it. Everything else runs in full on the free plan.
Prices in USD. Cancel from the billing screen at any time. No per-seat charge, no per-trace charge, no per-gigabyte charge.
What you get that the console cannot give you
OpenAI, Anthropic and Google bill you by organization, key and model, which are the dimensions of their system. Your questions are about yours.
| The question | Provider console | KostLens |
|---|---|---|
| Which customer of mine costs the most? | No | Yes |
| Which feature costs the most? | No | Yes |
| Every provider on one screen | No | Yes |
| Claude Code, Cursor and Copilot seats in the same total | No | Yes |
| What will this month's invoice be, across every provider? | No | Yes |
| One budget for every provider, not one per console | No | Yes |
| What would this month have cost on another model? | No | Yes |
| Which deploy made spend jump? | No | Yes |
| Who changed a budget or a rate, on any provider, and when? | No | Yes |
| The invoice you legally owe | Yes | No |
The last row is not a slip. KostLens does show every single call, with its cost, on a screen of its own. What it does not hold is the authority: your provider issues the invoice and applies your discounts and credits, and we read usage. When the two disagree, theirs is the one you pay. Plans are not metered by volume, though a fair-use ceiling on stored data does exist and is described in the terms.
Billing
No. Free is free forever, and it runs the product: all six screens, all three SDKs, budgets, alerts, the CI check and the MCP server. What it limits is capacity, one person and one project, and how far back you can look. The one thing that needs a paid plan is the activity log, because on a single-person account every line in it would name the person reading it. Nothing switches off on day 15.
Because you could not predict the bill before agreeing to it, and a company that sells cost predictability should not send surprise invoices. Capacity is something you can count in your head: how many people, how many projects, how far back. You know the number before you pay it.
No. On a paid plan your whole team is included at no extra cost, and so is every project. Per-seat pricing quietly punishes you for inviting the person who most needs to see the spend, which is usually whoever is paying for it.
The action is refused with a clear message, not silently dropped. Inviting a second person or creating a second project asks you to upgrade. Your data keeps being collected the whole time, so nothing is lost while you decide.
Nothing is deleted on the spot. The plan sets how far back you can READ, not how much we keep, and there is a grace period before the shorter window applies. Go back up and the older months are there again.
Closely, and rather than ask you to take that on faith the product measures it: the Overview puts our total for the month next to what the provider's own usage API reports, per provider, and names the gap. Cost is computed from token counts and published prices, versioned by effective date, including the parts nobody models: batch and flex halve the rate, priority and fast mode multiply it, and writing to cache costs more than reading from it. What we cannot see is your private discount or your credits, so the provider's invoice stays the source of truth for payment.
Yes, from the billing screen, and you keep the paid plan until the period you already paid for ends. No call, no email, no retention flow.
One person, one project, the whole product. Wired into your app before your coffee cools.
Start free