Quotas, limits and credits

Three levers keep AI cost in check: per-task rate caps, an optional per-customer daily cap, and moving usage onto customers' own provider keys.

Rate caps

RelyAI → Settings → Quotas & Limits. Each AI task has a rate cap so no one can hammer your provider.

Cap	Default
Per-task rate	20/minute, 500/hour
Per-customer daily	0 = uncapped

RelyAI → Settings → Quotas & Limits

Configure limits to prevent provider exhaustion:

• Per-task caps: Adjust per-minute and per-hour thresholds to balance speed and protection.
• Per-customer cap: Set a daily maximum to prevent individual accounts from monopolizing capacity.
• Provider backoff: Automatically handles upstream HTTP 429 responses with exponential backoff.

AI Credits

When plan-based AI metering is enabled:

• Customer dashboard: Displays an AI Credits card on the subscription page showing remaining allowance this cycle.
• Unmetered plans: Displays "No AI metering on this plan".
• Native accounting: Integrates directly with AcelleMail's built-in AI credit system.

Move usage onto customers' own keys

Offload API consumption to end-user accounts:

• Enable BYOK: Allow customer API keys under engine settings.
• User configuration: Customers enter their personal provider key under Account → AI.
• Quota isolation: All requests run against the customer's personal API quota instead of the system key.

Related pages

• Connect an engine
• Settings reference