Skip to main content

Privacy

This page explains what Broker as a Service (BaaS) collects when you use chainflip-broker.io, its API, its MCP server, the partner dashboard and this documentation site, and what happens to it. Questions or requests: info@chainflip-broker.io.

Your partner account​

When you register as a broker we store:

  • the partner name you register under,
  • your email address, if you give one,
  • your commission and the Ethereum address your commissions are paid to,
  • the API key and admin key we issue to you, and the broker addresses we create for you on the Chainflip network,
  • optional signup acquisition information: your answer to “How did you hear about us?”, a referral code, supported campaign labels from the signup link, and the referring website's domain (not its full address).

Dashboard users are stored with their email address, role, a password hash and, if you turn on two-step sign-in, an encrypted authenticator key and hashed recovery codes; we never store passwords themselves. We use your email address to send account emails: your keys at registration, sign-in codes, invitations, password resets, sign-in alerts, team changes, two-step sign-in changes and commission payouts.

We save acquisition information with your partner account to understand which sources lead to registrations and swap activation. We also use your registration time, the first successfully recorded usable quote since this measurement began, and recorded channel and swap-completion activity. A missing source is unknown, not proof of a direct visit; an earlier visit whose campaign labels do not reach signup cannot be reconstructed. Referral codes are information only and do not establish a reward or eligibility.

This partner-linked acquisition layer works on clearnet and Tor without attribution cookies, local/session storage, IndexedDB or persistent browser identifiers. It is separate from the anonymous usage statistics below: we do not send these account fields to Umami or join Umami sessions to partner accounts. Please do not put secrets or personal details in referral codes or campaign labels; format checks do not make text anonymous.

We retain this acquisition and activation information with the partner account until account-data correction or deletion through the contact process below; it has no automatic analytics expiry. Older copies can remain in backups until their retention expires. This does not change the retention of financial records or public blockchain data.

Swaps​

To quote, start and track a swap we store the assets, amounts, deposit and destination addresses and transaction hashes involved. All of these are public on the blockchains concerned anyway; a swap cannot be made private by us or anyone else.

We also store with each swap the IP address of the request that opened it, to investigate abuse and problems you report. Like the request log below, it is deleted after 15 days; the swap record itself stays. A swap opened over Tor stores only that it came over Tor, never an IP address.

Completed swaps may be announced in our public Discord and Telegram channels with the assets, amounts and their USD value, a link to the block explorer, and a partner label (your partner name, or a shortened broker address). Announcements never include wallet addresses, transaction hashes or email addresses.

Logs and security​

Requests to the API, the website and this documentation site pass through a request log we use for debugging. For each request it records the IP address, browser user agent and full URL, including the query string. We use it to investigate problems you report and to spot abuse. This log is kept for 15 days and then deleted.

Dashboard sign-ins and account changes are recorded separately, with the IP address and browser they came from, and are shown to you in your account's audit log.

Addresses that probe for well-known vulnerabilities are blocked for one hour. That block list is kept in memory only.

Our staff can open a partner's dashboard in read-only mode to help with a support request. Each such view lasts at most one hour and is recorded in our internal audit log.

Usage statistics​

To see which pages and tools are used, our servers count requests as they handle them and keep only daily totals. A page view is counted per day, site, page address (without its query string), the domain of the referring website (never its full address), whether it came over the clearnet or Tor, and a broad category guessed from the browser user agent: a person, a search engine, an AI crawler collecting training data, an AI assistant fetching a page for its user, or another automated client. For the MCP server we count, per day, successful initializations, tool calls and swaps started, with the tool used, whether an API key was given, and the name the MCP client application reports for itself, if any. Only a bounded number of distinct referring domains and client names is kept per day; the rest is counted as "other".

These daily totals contain no IP address, full user agent, query string, API key, email address or anything else that identifies you or your browser. This daily-count layer is collected on our server only, on both clearnet and Tor: no script runs in your browser and no cookie or other identifier is set for these counts. We keep the daily totals for years.

Clearnet browser statistics​

We use self-hosted Umami to measure visits to the clearnet landing page, signup page and documentation. It does not run on API or MCP endpoints, signed-in pages or Tor onion services. The script and event endpoint use the page's own origin; there is no hosted analytics provider.

This layer records the page hostname and canonical path, a referring website's DNS hostname (not its path or query), and only an approved utm_source campaign label from the page query string. Campaign labels must be public acquisition labels, never personal names or user identifiers: format checks cannot make personal text anonymous. Other query parameters, fragments, page titles and screen dimensions are excluded. It measures a signup-page view and three events: signup_submitted (a browser form-submission attempt, not successful account creation), docs_register_click, and docs_search_zero.

A settled zero-result documentation search may include its normalised search terms. Accepted terms are retained indefinitely with the event. Do not enter secrets or personal information into search. We reject common URL, email, wallet/key/token-like and overlong inputs, but these heuristics are not a guarantee that all personal or sensitive text will be removed; a rejected term can still produce a count-only event.

Umami processes the request's IP address and user agent transiently to derive browser, operating system and device categories, approximate GeoIP country, region and city where available, and anonymous server-side session identifiers with a daily rotating salt. There is no durable IP-address column in this analytics database. Cookieless does not mean no IP processing or no session identifier. We do not call identification APIs or join these sessions to accounts or partner records. Sessions and source-to-goal associations are approximate, not a deterministic cross-host or account funnel.

The integration honours Do Not Track and writes no analytics cookies, local/session storage, IndexedDB or persistent device identifier. Its collector cache token lives only in page memory. The upstream tracker may read an existing umami.disabled opt-out preference; this integration does not write it. Blocked or disabled JavaScript and Do Not Track visits are not measured by this layer. History has no automatic expiry: it is retained indefinitely until an operator deletes or resets it, and older copies can remain in backups until their retention expires. The separate 15-day raw request logs described above still include full requested URLs.

Tor​

Over our Tor onion services no IP address is recorded anywhere, in request logs, sign-ins, the audit log, swaps or the usage statistics (swaps and statistics only record that they came over Tor), and no third-party bot check is loaded. Tor hides who is calling the API, not the swap itself.

The clearnet browser-statistics layer does not run on onion pages; those pages do not load its script or contact a clearnet collector. The server-only daily counts still cover both transports.

Cookies​

The dashboard uses one session cookie to keep you signed in (one day, or 30 days if you choose to stay signed in) and a security cookie on forms. The public signup form also uses a necessary antiforgery (CSRF) cookie. These are not analytics cookies. The website and this documentation site use no advertising trackers, and neither statistics layer above sets analytics cookies.

Service providers​

  • Cloudflare Turnstile checks for bots on the sign-up, sign-in and password pages of the clearnet website. Cloudflare receives your IP address and browser information when it runs.
  • Discord and Telegram carry the public swap announcements described above.

Questions and requests​

To ask what we hold about you, or to have it corrected or deleted, write to info@chainflip-broker.io from the email address on your account. Swap data recorded on a blockchain cannot be deleted by anyone. See Contact Us for other ways to reach us.