Usage, Keys, And Billing
Top up non-expiring PAYG credits, optionally subscribe to Pro, and understand measured-energy billing.
PAYG And Pro Billing
PAYG has no monthly fee. Add $20, $50, $100, or $250 in usage credits and requests are charged at $10/kWh on Standard or $8.50/kWh on Flex, from measured energy when available.
Aegis Pro costs $10 per month and includes $5 of monthly usage credit, reserved concurrency, and advanced optimizer controls. The Pro credit expires at the end of its billing period; PAYG credits never expire.
Optimization is included for every paid user. When it reduces measured energy, the customer keeps the savings automatically; Aegis does not promise a fixed token-to-kWh reduction or pay a separate efficiency dividends.
Service Tiers: Standard And Flex
Every request runs on the customer-selected model. Choose Standard for the fastest response start, or Flex for a lower price in exchange for potentially slower request startup.
Standard: $10.00 per physical kWh. Best for interactive chat, live coding and debugging.
Flex: $8.50 per physical kWh. Best for autonomous coding agents, repository analysis, background research, code review and batch work.
Flex requires stream=true. A request that selects Flex without streaming is rejected with an OpenAI-compatible validation error (code flex_requires_streaming) and is never silently downgraded to Standard unless the customer explicitly sets aegis.allow_standard_fallback.
- Select the tier per request via the service_tier body field: "standard" or "flex".
- Omit service_tier to use the API key's default tier (Standard for existing keys).
- Override the default per API key from Console > API Keys > change default tier.
- Flex may delay request startup; Aegis does not guarantee a specific queue time.
API Key Lifecycle
Create keys from Console > API Keys. Keys are labeled, shown once at creation time, and can be deleted later from the same page.
- Use separate keys for local development, CI, and production agents.
- Rotate a key immediately if it lands in shell history, logs, or a shared paste.
- Deleting a key stops every client that still uses it.
Model Billing Behavior
Aegis keys are not model-locked. The client chooses the model on each request, and Aegis does not silently substitute a different customer-selected model. Billing primarily follows measured energy and uses a deterministic upstream-cost fallback when energy is unavailable.
Use the Models page to compare context limits and pricing characteristics before pinning one model in an agent config.
- Billable POST /v1 requests may start with any positive balance. Aegis atomically reserves the lesser of the available balance and the $0.25 per-request ceiling, so a $0.10 wallet funds one request limited to $0.10 while a $0.50 wallet can fund two $0.25 requests; GET /v1/models remains available for discovery.
- The exact dollar charge is reconciled after final usage arrives: unused reservation funds are released and measured energy or trustworthy upstream-cost evidence determines the actual debit.
- If a request reaches its actual reservation, Aegis cancels upstream. Streaming clients receive the billing-specific aegis_insufficient_credits event; partial-cap non-streaming requests are metered through an internal upstream stream and receive HTTP 402 without partial output when exhausted.
- Request IDs, charges, tokens, energy, refunds, and adjustments appear in the dashboard history.