Churnly data policy
Last updated 21 November 2026. This is Churnly's own statement of what it collects and what it does not. It is not legal advice. Churnly is run by Churnly. Questions or requests: hello@hichurnly.com. It sits beside our privacy page, terms and subprocessor list.
Who is responsible for what
- Your account and billing data: Churnly is the controller. This is what we need to sign you in, bill you, keep the service safe and meter the AI.
- Your customers' data: the customers, messages, notes, letters and settings you bring into Churnly are yours. You are the controller and Churnly is the processor, acting on your instructions. A data processing agreement is available at hello@hichurnly.com. Your own customers should send requests about their data to you; we will help you answer them.
- Early-access list: if you ask for access on hichurnly.com, Churnly is the controller of that address.
The short version
- We collect what is needed to run Churnly: customers, subscriptions, invoices and billing events read from your Stripe account; merged pull requests and releases from your GitHub repository; messages your customers send you by email or WhatsApp; and the saves, letters and replies Churnly drafts for you.
- We do not run analytics, advertising pixels or tracking cookies, we do not sell data, and we do not read or keep payment card numbers.
- Every account has its own database. Deleting your account in Settings removes it.
What we collect about you, the account owner
- Email address, and Google account ID if you sign in with Google. Source: you. Purpose: sign-in and account notices.
- Sign-in records. A hash of each sign-in link (valid for 15 minutes, once) and of each session token, with the time and the first 200 characters of your browser's user agent so you can see and end your sessions in Settings. We store no IP address.
- Plan and billing state. Plan, billing interval, status, the dates of the current period, a cancellation date, the date a payment first failed, your allowance cycle and top-up balance, and the IDs of your Stripe customer and subscription. Card details stay with Stripe.
- AI usage. For each AI call: time, the step, the model, token counts, cost and the number of messages used. Never the text.
- Audit log. A short line for account events: connecting or disconnecting a service (with the Gmail address, repository name or WhatsApp number involved), messages and replies sent, subscription and top-up events, and WhatsApp codes sent to your phone number.
- Your phone number, if you add it for WhatsApp alerts. The 6-digit code used to verify it is stored only as a hash and expires after 15 minutes.
- Settings you type: your name, business name, postal address, writing voice, time zone, discount limit, cadence and similar.
- Your edits to drafts: when you change a draft before sending it, the last 8 edits are kept as a before and after pair, each capped at 400 characters, and sent to the AI as examples of how you write.
- Style statistics: worked out each time from your sent messages and Sent-folder samples (typical length, greeting, sign-off, capitalisation). They are not stored, and are sent to the AI.
- Connection secrets: your Stripe key, Gmail refresh token, email app password, WhatsApp access token and GitHub token if you give one. Each is encrypted before it is stored and is never shown in the app or sent to the AI.
What we collect about your customers, and from where
- From your Stripe account (read every 2 minutes, with the restricted key you create): customers with their ID, name, email address and phone number; subscriptions with plan name, price, interval, quantity, status, start date, scheduled cancellation and the cancellation reason the customer chose; discount and promotion-code metadata used to find who referred them; a loyalty tier we store in the customer's metadata; and these events: failed and paid invoices, subscription changes, and checkouts left unpaid (with the email and name given and the amount). Purpose: spot who may leave and draft a save.
- Written to your Stripe account only after you approve: coupons, subscription pauses, payment retries, a credit to a customer's balance, the tier metadata, and card-update links.
- From your GitHub repository: the title, description (first 600 characters), link and date of merged pull requests and releases, read-only, to tell customers what shipped. Nothing is written to GitHub.
- From your customers' replies: by email through the reply address Churnly sets on your messages (routed by Cloudflare), or by your own inbox over IMAP if you connect it with an app password, or by WhatsApp. Over IMAP, every 10 minutes we read the sender, subject and date of recent inbox messages (30 days on the first run, up to 500 per pass) and download a message only when its sender matches a customer of yours. We keep the subject and text (first 4,000 characters, quoted history removed) of those; mail from anyone else is not stored, and automatic mail is ignored.
- From your Sent folder, only if you connect an app password: the text of up to your 10 most recent non-bulk sent messages (first 1,200 characters each), kept as samples of your voice, until replaced by newer samples or you delete your account.
- Created by Churnly: risk scores and reasons, signals, notes (a short quote and summary of what a customer asked for or reported), matches to what you shipped, proposed plays, drafts, letters, perks, and the log of each run.
- Per customer: an unsubscribe flag, a WhatsApp number once they write to you, and a private reply key used in reply and unsubscribe addresses.
What the AI sees
Drafts are written by Anthropic's Claude models through Anthropic's API. For each task we send only what that task needs: for a save run, the customer's name, email address, WhatsApp or phone number, plan, status, risk reasons, recent signals and up to the last 3 to 10 messages; for a reply, the customer's first name, plan, status and the last 8 messages; for letters, first names and short facts such as what they asked for and when they joined; for reading replies, subject and text; for matching, the notes and the titles and descriptions of what you shipped; plus your name, business name, voice, Sent-folder samples, your last 8 draft edits and style statistics. We never send access tokens or keys. Churnly does not store the prompts; it stores the drafts and run logs that come back. We do not train models on your data, and the Anthropic privacy terms for API use apply on their side. If AI is off or your allowance is used up, built-in templates write and nothing goes to Anthropic.
The AI cannot send anything. Every email, WhatsApp message, coupon, pause and retry waits for your approval. The one exception is auto-send, which is off until you switch it on and covers only plain thank-you replies with no link, price or promise that a second model has checked.
What we do not collect
- No analytics, advertising or tracking scripts, pixels or cookies on hichurnly.com or in Churnly. The marketing site sets no cookies and loads nothing from other companies.
- No payment card numbers, bank details or other payment credentials. Stripe holds them; we never read your customers' payment methods either.
- No reading of your mailbox through Google. Google sign-in asks only for your identity (openid, email); connecting Gmail sending adds gmail.send.
- No open or click tracking in the emails we send for you. They contain no tracking images and no rewritten links, and carry a one-click unsubscribe header and a "reply stop" line.
- No IP addresses stored. An address is held in memory for a short time to limit abuse and is passed to Cloudflare for the human check at sign-in.
- No contact lists. Customers come only from your Stripe account and the people who write to you.
- No passwords for Churnly: you sign in with an emailed link or with Google.
- No data sold, shared for advertising or used to train models.
Cookies and third-party scripts
Churnly sets one cookie that is needed to stay signed in (httpOnly, up to 30 days), and two that last 10 minutes during Google sign-in (a security check and the plan you picked). Two third-party scripts can load inside the app: Cloudflare Turnstile on the sign-in page, and Meta's script only when you connect WhatsApp. The marketing site's content security policy blocks all others.
Why we may process it
- Contract: your account, sign-in, billing and the service you asked for.
- Legitimate interests: security, rate limits, the audit log, preventing abuse and keeping AI cost under control.
- Consent: the early-access list, and each service you choose to connect, which you can disconnect at any time.
- Your customers' data: processed on your instructions. You are responsible for having a lawful basis to write to them.
Who processes it, and where
- Fly.io: hosting and storage in its US East region (Virginia), with daily volume snapshots.
- Anthropic: writes drafts and replies, and reads customer messages to note what they ask for.
- Stripe: our own billing, and the Stripe account you connect.
- Cloudflare: hosts hichurnly.com, stores the early-access list (address and time, in Cloudflare KV), routes inbound reply email and runs the human check at sign-in.
- Google: sign-in, and sending through Gmail when you connect it.
- Meta: WhatsApp messages when you connect it, and alerts and verification codes sent to your phone if you add it.
- Our email provider (named here before launch): sign-in links and account notices.
Services you connect also receive requests from us: your own mail server (SMTP and IMAP), and GitHub, which gets the repository name and your optional access token. Our hosting is in the United States, so data from elsewhere is transferred there; the data processing agreement covers this. The current list is on the subprocessors page.
How long we keep it
- Pruned automatically: messages after 24 months; runs and their event logs after 90 days; sessions after 30 days (or 7 days idle); sign-in links once they expire (15 minutes); short-lived notice markers after 30 days.
- Everything else is kept until you delete the customer, disconnect Stripe and choose to delete its customers, or delete your account: customer records, signals, notes, plays, drafts, letters, replies, perks, settings, Sent-folder samples, draft edits, AI usage records and the audit log.
- Early-access address: kept until Churnly opens and your invite is sent, then deleted, or removed sooner if you ask.
- Deleting a customer removes their records, and Churnly stops syncing that Stripe customer.
- Deleting your account in Settings cancels your plan at once, asks Google to revoke Gmail access if connected (deletion goes on if Google does not answer), stops our WhatsApp connection to your number, and deletes your database with every key and token in it, your sessions, sign-in links, audit log and account record.
- What remains after account deletion: Stripe's billing records for your Churnly plan, which Stripe keeps for tax; AI usage rows with the link to you removed (time, step, model, token counts, cost); changes already made in your own Stripe account and messages already sent; and Fly volume snapshots, which Fly takes daily and keeps 5 days by default.
How we protect it
- Connection secrets are encrypted with AES-256-GCM. Sign-in links, session tokens and phone codes are stored only as hashes.
- Each account has its own database file, and requests are scoped to the signed-in account.
- Sessions use httpOnly, secure cookies; cross-site requests are refused; sign-in, API and unsubscribe routes are rate limited; sign-in has a human check.
- Webhooks from Stripe, Meta and our inbound mail route are checked by signature before they are read.
- Mail servers you enter must be public hosts on standard ports.
- Text written by your customers is handled as quoted data in AI prompts, and the AI cannot send or change anything on its own.
- Pages are served over HTTPS with a content security policy. In production the service refuses to start without its secrets.
Your rights and how to use them
- Access and portability: Settings, Download my data, gives a JSON file of your customers, signals, actions, messages, replies, notes, shipped items, matches, letters, perks and settings.
- Deletion: delete one customer from their page, disconnect a service, or delete your account in Settings.
- Correction, objection, restriction, and the data the export leaves out (Sent-folder samples, draft edits, usage records, the audit log): email hello@hichurnly.com.
- Your customers: they can stop messages by replying "stop" or using the unsubscribe link, and can ask you to access or erase their data; you can erase them in Churnly.
- You may also complain to your data protection authority. California residents have the same access, deletion and no-sale rights. We do not sell or share personal information.
If something goes wrong
If we learn of a breach affecting your data, we will tell you without undue delay by email so you can meet your own duties to your customers. Report a security problem to hello@hichurnly.com.
Children
Churnly is a business tool. It is not for children and we do not knowingly collect their data.
Changes
We will update this page and its date when what we collect changes, and email you before a material change takes effect.