Legal / Terms of Service
Cent AI Terms of Service
Version 0.1 (draft) · Last updated: [TBD: effective date]
TOS-01 Who we are and what these Terms cover
- Cent AI ("Cent AI", "we", "us") is an independent service run as an individual business, not a company. You can reach us at support@cent-ai.dev.
- These Terms govern your use of the Cent AI website at
cent-ai.dev, the dashboard, the Playground, the documentation, and the API (together, the "Service"). - These Terms include the Acceptable Use Policy, the Refund & Credit Policy, the Privacy Policy, the Cookie Policy, and the Abuse Reporting & Takedown procedure. Links are in the site footer.
- By signing in or using the Service, you accept these Terms. If you do not accept them, do not use the Service.
TOS-02 Eligibility and accounts
- You must be at least 18 years old, or the age of legal majority where you live if higher, to create an account.
- You sign in with GitHub or Google. We accept only an email address that the sign-in provider has verified. For GitHub, this is your primary verified email.
- One account belongs to one individual. Team workspaces are not offered.
- You are responsible for all activity under your account and your API keys.
- Keep your sign-in accounts secure. If you think your account or a key is compromised, revoke the key and sign out all sessions from the dashboard, then contact us.
TOS-03 The Service
- Cent AI is a pay-as-you-go gateway that lets you reach many AI models through one endpoint and one API key.
- The API supports the OpenAI Chat Completions format, the OpenAI Responses format, and the Anthropic Messages format. Image, video and speech generation are marked "Coming soon" and are not available until we announce them; when available, TOS-10 and AUP-07 apply.
- Speech-to-text and embeddings are shown as "Coming soon". These endpoints currently return a "not available yet" error. We may open them later.
- The Playground lets you try models in your browser, including comparing up to 3 models side by side. Playground usage is billed from your balance like API usage.
- We pass your request parameters to the model provider as you send them. We do not change them silently. Parameters a provider does not support are returned as the provider's error.
- The model catalog can change. We may add models, remove models, or disable a model temporarily, for example when a provider is unavailable or its cost changes sharply. We will announce permanent removals in the public changelog in advance where we can.
TOS-04 Third-party model providers and sellers
- Models are served by third-party model providers and sellers, not by Cent AI. We route your request to one of them.
- The same model may be served by different providers at different times. We choose the route, including when a route fails.
- Your content is sent to the provider that serves your request so the model can process it. The provider's own terms and data practices also apply to that processing.
- Providers may apply their own safety systems. A provider may refuse or block a request. Cent AI does not add its own content filters to your requests.
- Model names and brands belong to their owners. We use them only to identify the models. Using a model through Cent AI does not create any relationship between you and the model's developer.
TOS-05 API keys, projects and management keys
- Keys have the format
cai_<secret>. The full key is shown once when you create it. We store only a hash and cannot show it again. - You can set optional limits on each key: a daily limit, a monthly limit, an IP allowlist, and a model allowlist. New keys have no limits by default. If you create a key without limits, any person who obtains it can spend your whole balance. That risk is yours. The dashboard warns you when you create a key without limits. A request over a key's daily or monthly limit is rejected with HTTP 429
key_limit_exceeded. A request from an IP address or for a model outside the key's allowlist is rejected with HTTP 403. - Every key belongs to exactly one project. New accounts get a "Default" project. You can create more projects and set an optional daily or monthly budget per project. When a project budget is exceeded, requests from that project's keys are rejected with HTTP 429
project_budget_exceededuntil the next period or until you raise the budget. Other projects are not affected. - Limits and budgets are checked when a request starts. Requests already in progress continue, so usage can go past a limit or budget. How far depends on how many requests were running.
- A management key has the format
cai_mgmt_<secret>. It can only read usage and balance data. It cannot make model requests. - You can create new keys and revoke any key at any time. Revocation takes effect within seconds.
- We do not scan public code repositories for leaked keys. Keep your keys private.
TOS-06 Balance, top-ups and payment
- Pay-as-you-go use of the Service is prepaid. Your balance is held in US dollars (USD). Budget packs are covered by TOS-06A.
- The minimum top-up is $1. The dashboard offers presets of $1, $5, $10, $25 and $50, or any amount from $1.
- You can top up in Indonesian rupiah (IDR) through QRIS on our payment processor's hosted page, using any banking or e-wallet app that can pay QRIS, or with USDT or USDC on the Solana network.
- For IDR payments, we convert USD to IDR at the live exchange rate, with no added spread. The rate is locked when you create the order and shown before you pay. An IDR payment order expires after 60 minutes.
- We pay the payment processing fees. You pay only the amount shown at checkout. Fees charged by your own bank, e-wallet or crypto wallet, such as blockchain network fees, are outside our control and are not covered.
- If you pay an IDR order after it expires, we still credit your balance automatically once the processor confirms the payment.
- For crypto, we credit the amount actually received after on-chain confirmation, including late deposits. If you send more than the order amount, we credit the full amount received. If you send less, we credit the amount received. USDT and USDC are credited at 1 token = $1. A deposit below $1 is also credited for the amount received; the $1 minimum applies only to creating an order. Funds sent on an unsupported network, in an unsupported token, or to a wrong address may be impossible to recover.
- We send a receipt by email with a PDF copy after each credited top-up. You can also download receipts from the Billing page. A receipt is not a tax invoice.
- Your balance is not a bank deposit, does not earn interest, and you cannot transfer it to another account. If you lose access to your account, we can move your balance once to a new account you own, after you show proof of payment and of the old email address.
- Your pay-as-you-go balance does not expire while your account exists.
- Prices and top-up amounts are the amounts you pay; we do not currently add VAT.
Example (illustrative numbers). You choose the $5 preset and pay with QRIS. Checkout shows $5.00, a locked rate of Rp16,500 per $1 and Rp82,500 to pay within 60 minutes. You pay Rp82,500. Your balance increases by $5.00. The processor's fee is paid by us. If you pay after 60 minutes, the order is still credited at the same locked rate.
TOS-06A Budget packs
- A budget pack is a one-time purchase of USD budget in three sizes: $1.99, $9.99 and $99. The budget you receive equals the price you pay. A pack is not a subscription: it does not renew and is never charged automatically. You can buy another pack at any time.
- Each pack size covers a set of models listed on the pricing page before you pay. Pack budget is used only for those models, at the prices in effect when each request is made. Other models are charged to your pay-as-you-go balance.
- Figures such as "up to N million tokens" are estimates based on the stated assumptions, not a guarantee. You are charged for actual usage.
- Each request is charged from your balances in this fixed order: promotional credit, then pack budget, then pay-as-you-go balance.
- When your pack budget runs out, requests continue from your pay-as-you-go balance if you have one; otherwise they are rejected.
- Pack budget does not expire while your account is active. Packs have the same rate limits as pay-as-you-go and get no priority.
- Pack budget is not cash and is not refundable, except when we correct a system error or a duplicate payment. If you delete your account, any remaining budget is not refunded.
- Pack prices are the amounts you pay; we do not currently add VAT.
- If you pay for a pack with crypto and send less than the pack price, the pack is activated with a budget equal to the amount received. If you send more, the full pack is activated and the excess is added to your pay-as-you-go balance.
TOS-06B Free daily event
We may offer a shared daily pool of free tokens on one model chosen at random each day. The pool is first come, first served, with no per-account allowance. After the pool runs out, requests to that model are charged at normal prices. We may change, pause, or end the event at any time. Free tokens have no cash value and cannot be transferred or refunded.
TOS-07 Pricing and charges
- Each chat model has four public rates in USD per 1 million tokens: input, cache read, cache write and output. The catalog also shows the official reference price and the actual discount per model. Image, video and speech models are priced as described in TOS-10.
- We charge each request using the rates of the price version active when the request was made.
- Cache rates apply only when the provider reports cache tokens for your request. We do not estimate or assume cache usage.
- If a provider does not report usage for a request, we estimate tokens with a tokenizer and mark that request as
estimatedin your Requests page. - Costs are calculated in units of $0.000000001 and rounded up per request.
- A minimum charge of $0.00002 applies to each successful request. We may set a different minimum for a specific model. Any model-specific minimum is shown on the model's page.
- Requests that fail before a provider accepts them are not charged.
- If a stream stops early for any reason, including when your client disconnects, you are charged for the tokens actually delivered.
- For chat completions, messages and responses requests, the cost of each request, in USD, is returned in the
usage.costfield of the response body and, for streams, in the final usage event. The same cost appears in your dashboard and exports. - Promotional credit is used before your pay-as-you-go balance.
Example (illustrative rates). A model's input rate is $0.50 per 1M tokens and its output rate is $2.00 per 1M tokens. You send 10 input tokens and receive 2 output tokens. The token cost is $0.000005 + $0.000004 = $0.000009. This is below the $0.00002 minimum, so you are charged $0.00002, and
usage.costshows0.00002. A larger request of 10,000 input and 1,000 output tokens costs $0.005 + $0.002 = $0.007, which is above the minimum, so you are charged $0.007.
TOS-08 Price changes
- Prices may go up or down at any time. A new price takes effect when it is published in our public changelog, which shows the model, the old and new rates, and the time it took effect.
- Each request is charged at the price in effect when we receive it. A price change never applies to requests received before it took effect.
- Budget you have already paid for keeps its dollar value. After a price change, the same budget may buy more or fewer tokens.
TOS-09 Promotional credit
- New accounts receive a one-time signup credit after signing in with a verified email: $1 for the first 100 accounts, then $0.10. One credit per account, valid for every model for 30 days.
- If someone you refer makes a first top-up of at least $1 and it is credited, you receive promotional credit equal to 10% of that top-up. Each referred user can earn you a reward once. You can earn at most 20 referral rewards per month. Self-referrals are not allowed. The reward is granted only after the referred user's top-up is final, and it is not taken back if that top-up is later disputed or reversed.
- Promotional credit can be used on all models, expires 30 days after it is granted, cannot be withdrawn or refunded, and cannot be transferred.
- We email you 7 days and 1 day before your promotional credit expires, unless it is already used up.
- We may remove promotional credit obtained through abuse, such as multiple accounts or self-referrals.
TOS-10 Media generation charges (images, video and speech)
This section applies once media generation is available to your account.
- Media models are priced per variant, for example per image at a given resolution or per second of video at a given resolution and with or without audio. Each model page lists every variant with its price, the official reference price and the discount. When a provider charges a media model by tokens, the page shows the token rates and an approximate price per image or per second, with the assumption used (a 1024×1024 image or a 5-second 720p video). You are charged at the listed price of the variant you use, or at the listed token rates. The $0.00002 minimum also applies.
- Before a media request runs, we show an estimated cost and hold that amount from your available balance. The estimate is an upper bound based on your request parameters, such as size, quality, duration and number of outputs. If your available balance is lower than the estimate, the request is rejected and not sent to a provider.
- When the request finishes, we charge the actual cost and release the rest of the hold.
- If the media request fails or is rejected, including after the provider has processed it, you are not charged and the full hold is released. If you request several images and only some succeed, you are charged only for the ones that succeed.
- Generated images, video, audio and related uploads are stored for up to 24 hours and then deleted automatically. Download anything you want to keep.
Example (illustrative numbers). You ask for 4 images of a variant priced at $0.010 per image. The estimate is $0.040, so $0.040 is held. Your available balance drops by $0.040 while the request runs. All 4 images succeed: we charge $0.040. If instead 2 of the 4 images fail, we charge 2 × $0.010 = $0.020 and release $0.020. If all 4 fail, we charge $0 and release $0.040.
TOS-11 Usage limits
- Each API key is limited to 600 requests per minute. Playground usage has the same limit. Limit headers are returned with each response. You can ask us for a higher limit.
- When your balance is zero or below, new requests are rejected with HTTP 402
insufficient_balanceand are not sent to a provider. Requests already in progress continue and are charged, so your balance can become negative. How negative depends on how many requests were running. A negative balance is offset when your next top-up is credited. If an account with a negative balance is deleted or blocked, we write off the negative balance and do not collect it. An account blocked after a payment reversal is not unblocked until the reversed amount is paid (RCP-03.7). - A single stream can run for up to 2 hours by default. Some models may have a different maximum.
- We do not send low-balance emails. The dashboard shows a banner when your balance runs out.
TOS-12 Availability
- We aim for 99.5% monthly availability of the API. This is an internal objective, not a contractual service level. We do not offer service credits.
- We publish a status page and per-model status (latency, time to first token and error rate) based on real traffic.
- We may perform maintenance, and providers may become unavailable. We will report incidents on the status page.
TOS-13 Your content and outputs
- "Content" means the prompts, inputs, files and other data you send, and the outputs you receive.
- As between you and Cent AI, you keep your rights in your inputs. Cent AI claims no ownership of your inputs or outputs. Rights in outputs may also be subject to the terms of the provider that served the request.
- You give us permission to process, transmit and, where described in the Privacy Policy, temporarily store your content only to provide the Service to you.
- We do not use your content to train models and we do not sell it.
- You are responsible for having the rights and legal basis to send your content, including any personal data of other people. If you build an application on the Service, you are responsible for your end users' use of it.
- AI outputs can be wrong, incomplete or offensive. Review outputs before you rely on them.
TOS-14 Data and privacy
- By default we do not store the content of requests that succeed. We always store request metadata, such as time, model, tokens, cost and status, to bill you and show your usage.
- When an API request fails, we keep an encrypted error record with its full content, including the key you sent, for 30 days to diagnose the failure. It is readable only by Cent AI operators and is never used for training.
- You can turn on "store request content for 24 hours" for a specific key. Stored content is deleted automatically after 24 hours.
- Details are in the Privacy Policy (PRV-06).
TOS-15 Acceptable use and suspension
- You must follow the Acceptable Use Policy.
- We may suspend or block your account or keys, without prior notice where needed, if we reasonably believe you violated these Terms or the Acceptable Use Policy, if your use creates a security or legal risk, or if a provider or authority requires it.
- While your account is suspended, your keys stop working and you cannot spend your balance. Your balance is not deleted. Promotional credit still expires on its original date.
- We will tell you the reason for a suspension and how to appeal, unless the law or an investigation prevents us.
- If your account is blocked permanently, your balance is frozen: it cannot be used, withdrawn or refunded, and it is not deleted. If the block is lifted, you can use the balance again.
TOS-16 Account deletion
- You can delete your account at any time in Settings.
- Before deletion, you can export your data as CSV.
- Deleting your account revokes all keys and sessions and anonymizes your personal data. Your name is removed and your email is replaced with a one-way hash, which we keep only for the purposes in the Privacy Policy (PRV-13).
- Any remaining balance, including promotional credit, is not refunded when you delete your account.
- We keep financial records (ledger entries) without your personal data to meet accounting obligations.
TOS-17 Refunds
Refunds and credits follow the Refund & Credit Policy. Budget packs follow TOS-06A and RCP-12.
TOS-18 Our intellectual property
- The Service, the Cent AI name and logo, the website, the documentation and our software belong to us or our licensors.
- You may use the documentation and code samples to integrate with the Service.
TOS-19 Disclaimers
- The Service is provided "as is" and "as available".
- To the extent the law allows, we disclaim all implied warranties, including fitness for a particular purpose and non-infringement.
- We do not guarantee that any model, provider, price or discount will remain available.
TOS-20 Limitation of liability
- To the extent the law allows, we are not liable for indirect, incidental, special or consequential damages, or for lost profits, revenue or data.
- To the extent the law allows, our total liability for all claims relating to the Service is limited to the total top-ups you paid in the 3 months before the event giving rise to the claim.
- Nothing in these Terms limits liability that cannot be limited under applicable law.
TOS-21 Indemnity
You agree to compensate us for claims and losses caused by your content, your application, or your violation of these Terms or the law.
TOS-22 Changes to the Service and these Terms
- We may change the Service. Price changes follow TOS-08.
- We will notify you of material changes to these Terms by email and a dashboard notice at least 14 days before they take effect.
- If you keep using the Service after a change takes effect, you accept the updated Terms. If you do not accept them, delete your account before that date.
TOS-23 Disputes
Before taking any formal step, contact us at support@cent-ai.dev so we can try to resolve the issue informally within 30 days.
TOS-24 General
- These Terms are available in English and Indonesian. If the versions differ, the English version prevails for all users. The Indonesian version is provided as a full translation.
- If any provision is found invalid, the rest stays in effect.
- We are not responsible for delays or failures caused by events beyond our reasonable control, including provider outages, network failures and government actions.
- You may not transfer your account or these Terms. We may transfer them as part of moving the business to a legal entity, with notice to you.
- These Terms, together with the policies listed in TOS-01, are the entire agreement between you and us about the Service.
TOS-25 Contact
- Support: support@cent-ai.dev.
- Legal notices: support@cent-ai.dev. Messages in chat channels are not legal notices.