Privacy Policy

FluxRouter is operated by Ferrox Labs, LLC, a Florida limited liability company. FluxRouter is an API gateway that routes your AI requests to underlying model providers. This page explains, in plain language, what data we handle and how. For questions or data requests, email support at fluxrouter dot ai.

What we collect

• Account information — your email and the details needed to sign you in and run your account.
• Usage and billing data — request metadata such as token counts, cost, latency, and which model served the request, so we can route, meter, and bill accurately.
• Payment information — handled by our payment processor, Stripe. We do not store your full card number.
• Support messages — if you contact support or use our in-page support assistant, we process what you send us to answer you.
• Flux Memory data— only if Flux Memory is on for your key, the text of the messages you send and the replies you receive on that key, and a search index built from them. See "Flux Memory" below.

How we use your data

We use it to provide and route the service, meter usage and bill you, support your account, keep the service secure, and meet our legal obligations. Where the GDPR applies, our legal bases are: performance of our contract with you (to provide and bill the service), our legitimate interests (to secure and improve the service), and compliance with legal obligations.

Your prompts and responses

FluxRouter is a routing gateway, not a model trainer. When you make a request, its content is forwarded to the underlying provider that serves it, and that provider processes it under its own terms. To operate, secure, and protect the service, we keep a PII-redacted (scrubbed) copy of the request for up to 30 days, then delete it. Outside Flux Memory, we never keep the raw, unredacted content. When Flux Memory is on for a key, we keep the text of that key's messages and replies as described under "Flux Memory" below. We keep request metadata that contains no prompt content (such as token counts, cost, latency and the model selected) for longer, for billing and analytics. We do not use your prompts or responses to train any FluxRouter model; our internal routing-quality systems are trained only on synthetic, internal traffic — never on your content. We do not sell or share your personal information (as those terms are defined under applicable law, including for cross-context behavioral advertising).

Flux Memory

Flux Memory is an optional feature that remembers your conversations, so later requests can use what you said earlier. It is off unless it has been turned on for your key. When it is on, it works on Chat Completions requests (/v1/chat/completions) only.

What we store. The text of each message you send and each reply you receive on that key (image, audio and file parts are not stored); a search index built from that text; and counts and timestamps used for limits and usage. We do not redact this text, so it contains whatever you or your users typed, including any personal information.

Why. To add relevant earlier passages to your later requests and to measure usage. We do not use memory to train any model, and we do not sell or share it.

Who processes it. We build and search the index inside our own memory service; no outside provider receives your memory for that. When a passage is added to a request, the model provider serving that request receives it along with the rest of the request. The memory database runs on DigitalOcean in New York, USA. Each memory is encrypted at rest with its own AES-256-GCM key, and those keys are protected by Amazon Web Services Key Management Service in the US. The service must decrypt memory to search it, so authorised FluxRouter staff could technically access it; access to production systems is restricted and logged.

How long we keep it. Until you delete it.

Turning it off. Send x-flux-memory-capture: off to keep a request out of memory, x-flux-memory-inject: off to stop memory being added to a request, or x-flux-no-retain: 1 to turn off memory and all other retention for that request. Requests sent with store: false are not remembered.

Deleting it. Email support at fluxrouter dot aito delete all or part of your memory, or one of your end users' memory. Deleted content stops being used at once and is erased from our live storage within 24 hours by destroying its encryption key. Backups expire within 7 days; we keep a record of each deletion (without content) outside the database and re-apply it before any restored backup is used. Text already sent to a model provider in a request is governed by that provider's terms.

Your end users. If your key is set up for many users and you send an end-user id, we keep a separate memory for each id. A key set up as personal has one memory, whatever id is sent. We cannot verify who a person is; we separate memory by the id your systems send. Send ids only from your own signed-in sessions, and tell your users how their data is handled.

Bot email

Bot email is optional. It gives a bot you run its own email address. When the bot sends or receives mail, we handle that mail for you.

People who email a bot.If you write to a bot address, or a bot writes to you, we process your email address, your name if it is in the message, the subject, the message, any attachments and the standard mail headers. You are not a Flux customer, and we use this only to deliver the mail and to run the service for the bot's owner. We scan incoming mail for spam and for hidden instructions aimed at the bot. We do not sell this data, and we do not use it to train any model.

Who processes it. Our email provider, AgentMail, delivers the mail, stores it and runs the scan. Mail is stored in the United States.

How long we keep it.

• Mail and attachments stay while the inbox is open. When the owner removes an inbox, it closes for 30 days so they can bring it back. When those 30 days end, we delete the mail.
• A message waiting for the owner's approval is kept until it is approved, discarded or 72 hours pass. Then we delete it.
• Event records (ids, sender, subject and a short preview) are kept for 7 days.
• Sending records (a scrambled form of each recipient's address, counts, cost and outcome) are kept for 13 months.
• If you complain, reply STOP, or an address bounces, we keep a scrambled, one-way form of your address, not the address itself, for 13 months. This stops any bot on the service from emailing you again. Removing an inbox or an account does not remove it.
• A bot address that has been removed is retired for good. No one can use it again.

Asking us to erase your address. If you are not a Flux customer and want your address erased from bot email, email support at fluxrouter dot ai and tell us the address. We delete it from our event records, held messages, approval records and recipient lists. We ask our email provider to delete its copies. We keep only the scrambled stop-list form, so that you are not emailed again. We reply within 30 days.

Service providers we rely on

To run FluxRouter we use a set of trusted providers, each of which handles data under its own terms:

• Stripe — payments and billing.
• Supabase — authentication and database.
• Cloudflare — security and content delivery.
• Resend — transactional email.
• Chatbase — our in-page support assistant; it processes the messages you type into it.
• DigitalOcean — hosting, including the Flux Memory database.
• AgentMail: delivers, stores and scans bot email.
• Amazon Web Services — encryption key management, and a record of Flux Memory deletions that contains no content.
• AI model providers — the third parties that serve your requests.

Where your data is processed

We process data in the United States and may use infrastructure in other regions, and your request content is processed wherever the provider that serves it operates. If you are in the European Economic Area or the UK, this means your data may be transferred internationally; where required, we rely on appropriate safeguards such as Standard Contractual Clauses. Flux Memory data is stored in the United States (New York). Bot email is stored in the United States.

How long we keep data

We keep account data while your account is active, and billing and tax records for as long as the law requires. We delete the scrubbed copy of each request after 30 days. We keep request metadata that contains no prompt content (such as model, token counts, cost and timestamps) for 13 months, then delete it. We keep Flux Memory data until you delete it. We keep bot email data as described under "Bot email". We delete or de-identify other data when it is no longer needed for these purposes.

Your rights and choices

You can request access to, correction of, or deletion of your account data by contacting us. You can also ask us to delete or export your Flux Memory data. If you are not a customer but a bot has your email address, see "Bot email" for how to have it erased. Depending on where you live, you may have additional rights — including to know what we hold, to delete, to correct, to data portability, to object to or restrict certain processing, and to opt out of any sale or sharing of personal information (we do not sell or share). We will honor verified requests except where we must retain records for billing, legal, security, or fraud-prevention reasons, or for de-identified, aggregate data. EEA/UK users may also lodge a complaint with their supervisory authority. Keep your API key secret; it is tied to your account and billing.

Children's data

FluxRouter is not directed to children and is intended for users 18 and older. We do not knowingly collect personal information from children. If you believe a child has provided us data, contact us and we will delete it.

Security

We use reasonable measures to protect your data, including encryption in transit, access controls, and vetted service providers. No system is perfectly secure. If a breach affects your personal information, we will notify you and the authorities as required by law.

Contact

For any privacy or data-handling question, email support at fluxrouter dot ai. See also our Cookie Policy. We may update this policy from time to time; the date above shows the latest version.