How credits work
The Quant Agent doesn't run on your laptop — it runs on FreqEdge's secure AI gateway, powered by frontier language models. Each operation costs a fraction of a dollar, drawn from your credit balance. One credit = $1, and your balance is shown in dollars in your Console. Credits are shared across all bots in your account, and are deducted only after an operation completes successfully.
Every paid tier includes a monthly Copilot credit allowance — $2.50/mo on Starter, $10/mo on Pro, $40/mo on Scale — granted automatically at the start of each billing month, no action needed. Once the monthly allowance runs out, the Copilot pauses until you top up; your bots keep trading unaffected.
What each action costs
| Operation | Cost |
|---|---|
Draft from spec Turn a written rule-set into draft Freqtrade code, validated in the sandbox. Saved as a draft; you backtest and promote it. | $0.50 |
Debug Diagnose a strategy error or failed backtest and suggest a fix you can apply as a new version. | $0.50 |
Narrate Analyze a backtest result with risk metrics, pair attribution, and plain-English recommendations. | $0.25 |
Regime Read the current market regime across your pairs and say which of your bots it suits. | $0.35 |
Agent step One turn of a validation or experimental run. Does not deploy. Does not guarantee improvement. | $0.20 |
Costs are fixed per action. You're charged only when an operation completes successfully — failed or timed-out calls cost nothing.
Topping up
Every paid tier includes Quant Agent skills — you simply top up credits as you go. Add credits any time from the Credits page in your Console, in $5, $10, $25, $50, or $100 packs. Purchased credits are shared across every bot in your account.
Why metered AI?
Transparency
You see exactly what each action costs before you run it, and every deduction is logged in your Console with a timestamp and operation type. No mystery bills.
Privacy
Your exchange keys never touch the AI gateway. Only the strategy code and backtest metrics for the action you choose are ever sent — the zero-knowledge architecture is unchanged.
Sustainability
Frontier AI models are powerful but expensive. Metering keeps the platform alive without forcing light users to subsidize heavy ones — you pay only for what you use.