Wallets and access
Free access and Paid-mode billing
Adding funds does not start billing automatically. Your account stays in Free mode until you explicitly enable Paid mode in the dashboard.
Usage mode
You control when wallet funds are available
Free mode
Free models remain available under the Free allowance. Premium-model requests are rejected before quota or wallet funds are reserved, so deposited funds remain untouched.
Paid mode
Premium models are enabled and use the Paid allowance. Only premium-model requests reserve and settle wallet funds. Free models stay on the Free allowance without wallet billing.
The mode is account-wide and applies to every API key. Switching modes does not reset request or token counters. Paid mode requires a positive available balance.
Free models
Quota controlled
Non-premium models never debit the wallet. They consume the account's Free-model RPM, TPM, daily request, and daily token allowance even while Paid mode is enabled.
Paid mode
Premium requests are wallet controlled
For a premium model, Edy reserves the maximum estimated cost before contacting the provider. A wallet without sufficient funds receives HTTP 402. Free-model requests do not create wallet reservations.
How settlement works
- 1. Input tokens and requested maximum output are estimated.
- 2. Premium requests reserve that maximum cost atomically.
- 3. The provider reports actual input, cached input, and output usage.
- 4. Edy deducts the actual cost and returns the unused reservation.
Prompt caching
Cached input is billed at a lower rate
When the upstream provider returns prompt caching metrics, Edy separates the billed input into two portions: fresh input tokens charged at the model's standard input price, and cached input tokens charged at the model's cache read price. Each model catalog entry has three independently configured price fields:
Input price / 1M
Rate applied to fresh, uncached prompt tokens per million tokens.
Output price / 1M
Rate applied to completion tokens per million tokens.
Cache read price / 1M
Reduced rate applied to cached prompt tokens per million tokens.
The total cost is calculated as (fresh input × input price) + (cached input × cache read price) + (output × output price). If a provider does not report cached tokens for a request, all input is billed at the standard input price.
Changing modes
Open the dashboard and use the Free/Paid control under your wallet balance. A premium request made in Free mode returns HTTP 403 with the code paid_mode_required. Switching back to Free blocks premium requests. Free models remain free in either mode.