Credits, Quotas & Margins

Omni utilizes an authoritative credit ledger to track LLM usage, tool execution costs, and rate quotas.

Two-Tier Quota Architecture

1. Synchronous Quota Admission

Before admitting a billable agent run or completion request, the edge proxy verifies account standing and reserves quota atomically. If the user has exceeded their rolling limit, the request is rejected with 429 Quota Exceeded before incurring upstream LLM provider costs.

2. Post-Execution Token Settlement

Upon successful run completion, exact prompt and completion tokens are recorded in the central usage ledger. If a run fails prior to model invocation, the reserved quota is immediately released.

Rolling Windows & Guardrails

Omni enforces rolling window checks:

• 1-Hour Window: Protects against rapid burst exhaustion.
• 5-Hour Window: Prevents continuous automated drainage.
• Monthly Token Ceiling: Overall account entitlement ceiling.

Checking Balances via API

Query your current usage statistics and remaining quota:

curl -sS "https://edge.omnistatic.com/v1/credits" \
  -H "Authorization: Bearer $OMNI_API_KEY"

Response:

{
  "used": 142500,
  "limit": 1000000,
  "remaining": 857500,
  "month": "2026-08"
}