AI Usage and Credit Policy

IrisOS, available at https://irisos.app, is operated by GRL Enterprises (we, us, or our). Our business address is 307, Grima Unnathi, Kyalasanahalli, Bengaluru – 560077, Karnataka, India.

This policy supplements the Terms of Service. It describes the reviewed implementation; it does not authorize a new price, credit purchase or recurring charge.

• Terms of Service
• Privacy Policy
• Cancellation & Refunds

Allowances are shared by the Workspace’s users. Reaching either the input or output limit can stop a request even when the total limit has not been reached. There is no automatic paid overage purchase. Provider allocation counters do not impose separate hard provider quotas in the reviewed reservation logic.

• Basic (shown as Free): 20,000 total credits, including at most 6,000 input and 14,000 output credits per calendar month.
• Pro: 200,000 total credits, including at most 60,000 input and 140,000 output credits per calendar month.
• Advanced: 500,000 total credits, including at most 150,000 input and 350,000 output credits per calendar month.

The system ordinarily counts reported input and output tokens with equal unit weight. It does not implement a separate provider-price multiplier or cached-token discount. Different providers may tokenize content differently. An IrisOS credit is therefore not a promise of exactly one raw model token for every feature.

Input can include internal instructions, the routing prompt, your prompt, history and supplied workspace context. Output includes generated responses and model-generated workflow material. Where usage is unavailable, text helpers can estimate usage from UTF-8 byte counts. Transcription uses provider usage when available and otherwise estimates input from audio size and output from transcript length.

A request can invoke classification, planning and completion separately; each metered call can consume allowance. Metered automatic summaries and background/workflow calls also count. Merely executing a non-AI tool is not separately assigned a credit price, but surrounding AI calls may consume credits.

Most server AI calls reserve estimated input and a maximum output allowance before sending the provider request. This can reject a request whose reservation exceeds remaining input or output capacity even if its final response would have been smaller. Settlement reduces the reservation to measured or estimated usage, capped at the reserved amount.

A provider call that fails and is released by its error handler normally returns that call’s reservation. A successfully completed routing or helper call can still count when a later stage fails or produces an unsatisfactory result. A partial stream, lost connection or process failure may require reconciliation; we do not promise that every incomplete user request consumes zero credits.

Realtime voice usage is recorded from usage events reported by the voice client after processing. It does not follow the same pre-reservation path as text. Report unexpected deductions, repeated-event charges or reservations that remain after failure for investigation.

Allowance counters reset for each calendar month at 00:00 UTC on the first day; the reset is applied when usage is next checked. This is independent of a purchase or billing-renewal date. Unused allocation does not roll over.

Upgrading changes available limits but does not erase the month’s prior usage. Cancelling immediately applies Free limits while preserving usage; if usage already exceeds those limits, requests may be unavailable until the next reset or an eligible upgrade. Historical usage events can remain after a counter reset.

Miles can automatically route among configured OpenAI, Google Gemini and Anthropic Claude integrations. A query may be sent to Gemini for classification before another provider completes it. Availability, functionality, performance, cost and routing rules can affect provider selection; model substitution does not by itself authorize an undisclosed credit multiplier.

Providers have different underlying costs, but the current accounting rules above apply. A material change to credit valuation or pricing requires appropriate prospective disclosure and notice under the agreement and law. Your existing usage cannot be repriced retrospectively.

• Provider Register
• Privacy Policy

No separate AI-credit top-up purchase, balance-expiry or refund mechanism was established in the reviewed implementation. Future paid add-ons must disclose their terms before purchase. Monthly unused allowance has no independent cash redemption value; statutory rights and express order terms remain intact.

For duplicate deductions, suspected metering errors or qualifying service failures, contact support with the workspace, approximate request time and relevant usage details. Do not include confidential prompts unless necessary. Confirmed errors should be corrected; the cancellation policy governs monetary refunds and mandatory remedies. Provider outages do not create an automatic service-credit entitlement beyond an accepted agreement or law.

• Cancellation & Refunds

The usage service provides Workspace totals, input/output usage, provider totals and reset timing. Access to detailed usage history depends on the interface and role; a complete per-request provider trace is not promised. Do not evade quotas, tamper with usage reporting or consume another Workspace’s allowance.

For privacy, legal, billing or service requests, email contact@irisos.app, use Get Help in IrisOS, or write to GRL Enterprises at the business address above. Include the relevant workspace and request details; do not send passwords or payment-card credentials.

Material changes will be communicated as required by the applicable agreement and law, with a prospective effective date. Where renewed consent or acceptance is required, it must be obtained before the change applies. A revised policy does not retroactively authorize a new use of personal data or impose new charges.