Your work does not pause when a weekly counter resets.
If you run coding agents on a subscription, you know the pattern: a weekly window, a counter, and the choice between stopping mid-task or working strange hours to spend the allowance before it vanishes.
How the other meter works
Subscription plans meter usage in weekly windows that reset on a schedule. Heavy agent weeks hit the ceiling mid-task, long runs stall at the cap, and the schedule of your work gets set by the schedule of the reset.
How metering works here
There are no limit windows. Platform usage bills transparent credits at the prices shown in the app. Your own API keys and local models are free lanes and always uncapped. A heavy week costs what it costs; it does not stop when a counter resets.
The real math
A subscription trades a fixed bill for a fixed ceiling; that is a fair deal while your usage fits the window. When it does not, the ceiling becomes the constraint on your work. Here, you pay for what runs, at per-model prices displayed before you run it, with a top-up fee (5%, $0.50 min) on credit purchases. Cheap open models keep routine work inexpensive; the frontier model is a dropdown away for the tasks that need it.
Compare the two meters yourself: idapt vs Claude Code and idapt vs OpenAI Codex.
What de-risks the switch
- Start free, no card: your keys and local models work free from minute one.
- You set your own spend caps; you are never charged past your plan.
- Credits never expire, and unspent credits are refundable on request.
- Your first subscription charge is refundable in full within 14 days.
Keep working past the reset
Start free, no card. Bring your own keys, run local models, or add credits when you want the frontier models.
Start free