Privacy policy

Last updated 6 November 2026

We take your training seriously and your data more seriously. This page says what Almost Fire collects, why, who else touches it, and how to get rid of it. Almost Fire is run by David Gibbon in Australia. Questions go to hello@almostfire.com.

What we collect

Your account: your name, email address, and a password hash, or the account ID Apple or Google gives us when you sign in with them. With Apple we also keep, encrypted, the token Apple gives us, so that deleting your account also ends Almost Fire’s link to your Apple ID.
Your training: your program, sessions, sets, loads, reps, effort, notes, aims and check-ins, including any aches, pain or symptoms you tell us about, and what you tell us when you set up your plan (goal, experience, body weight, training days, areas to protect, and whether a doctor has limited or cleared you).
Your food: the meals, foods, barcodes, macros and body weight you log.
Photos: meal, food label, barcode, equipment, habit and chat photos you choose to take or send. They are stored with your account so Stoke and you can see them again. On an iPhone with LiDAR, a food photo also carries a depth map of the plate, which we use to measure how much is on it.
Food snaps we keep to improve measuring: each food snap (the photo, its depth map when there is one, and what the app measured and read from it) is kept with your account so we can check and improve how the app weighs food. It is never used for advertising, and it is deleted with your account.
Your conversations with Stoke: your messages and Stoke’s replies. When you dictate, or talk to Stoke on a live call, your voice is streamed to our AI provider to be understood and answered; we keep the text, not the audio.
Habits and check-ins: the habits you track and tick (with a photo when that is the proof), your morning check-in and how each session felt.
Apple Health, when you connect it: your cardio workouts (time, distance, heart rate and energy), and a daily summary of your sleep, resting heart rate, heart rate variability and cardio fitness (VO2 max). They are stored with your account to count your runs and rides in your plan and to draw your graphs on Progress. Stoke reads them only if you turn that on, and you can turn it off at any time in Settings. A sleep habit can tick itself from them. Health data is never used for advertising, marketing or data mining, never sold, and never stored in iCloud.
App analytics and crash reports: which screens you open, which features you use, errors and crashes, your device model and app version, and when you open and leave the app, linked to your account ID. We also keep masked session replays, a recording of how screens move and where you tap with every word, number and image hidden, to see what led up to an error. We don’t store your IP address or location.

What your crew sees

A crew is up to six people you invite yourself. Nothing is shared until you join one. While you’re in a crew, its members see your first name, each session you finish in the last four weeks (its name, time, volume, sets, bests, exercises and the loads and reps you lifted, plus its score and Stoke’s one-line read), how many of this week’s planned sessions you’ve done, the cheers you send, the captions, comments and photos you choose to post, the name and weekly result of any habit you choose to share with them, and a note when you reach a run of morning check-ins. They never see your check-in answers, aches or symptoms, food, body weight, notes, Apple Health numbers, other photos or Stoke conversations. On a day you report pain, they see the session without its score or Stoke’s read. Leave the crew, or be removed, and your posts and cheers stop showing to it straight away.

Before a caption, comment, shared habit name or photo goes to your crew, an automatic filter (OpenAI’s moderation service) checks it for objectionable content. This check is part of keeping crews safe, so it runs for everyone in a crew.

If you report a crew-mate’s post, comment or the crew-mate, we keep a copy of what you reported (their first name, the words and how many photos) with your name, so we can review it. If you block someone, we keep that you blocked them so the app can keep you apart. Both are deleted with either account.

What stays on your phone

Apple Health during a session: Almost Fire saves each session to Apple Health as a workout, and reads your heart rate and active energy during a session to show them on screen.
Calendar and reminders: sessions are added to your calendar when you ask, and session reminders are scheduled on your device.

Why we use it

To run the app: build and adjust your program, log your sessions and food, let Stoke coach you, keep your account working, and fix what breaks. We use analytics to learn which parts of the app help and which don’t. We don’t sell your data, we don’t show ads, and we don’t track you across other companies’ apps or websites.

Who else processes it

A few services do work for us. They process your data only to provide that service:

OpenAI, Anthropic and Google (Gemini) run Stoke, our AI coach. The app asks your permission before anything goes to them. With it, your messages, voice, the photos you send, and the training, food and check-in details Stoke needs to answer you, review a session or write your weekly review go to them; your Apple Health numbers go too only if you also turn that on. Google’s Gemini reads food photos to recognise foods you have logged before. Under their API terms they don’t use it to train their models. You can take the permission back at any time in Settings › Stoke and AI: the app keeps working, without Stoke.
ChatGPT (OpenAI) and Claude (Anthropic), when you connect Almost Fire to them yourself. Nothing goes until you sign in to Almost Fire from that app and allow it. After that, what you ask for there (your plan, sessions, sets, food, weigh-ins and progress, and the meal photos you send) goes to that app to answer you, under your account and terms with them. Your Apple Health numbers never go. Disconnect in that app at any time.
OpenAI’s moderation service checks what you post to a crew, as above.
Vercel (hosting) and Neon (database) store and serve your account and training data.
PostHog receives the analytics and crash reports above.
Resend sends account emails, such as password resets.
Expo delivers app updates and crew notifications (it receives your device’s push token and the notification’s words).
Open Food Facts looks up a barcode when you scan one. Only the barcode is sent.

Each of these providers is bound by its terms with us to protect your data at least as well as this policy does, and to use it only to provide its service. Some are outside Australia, mostly in the United States, so your data may be stored or processed there.

Keeping and deleting it

We keep your data while you have an account. Delete it any time in the app under Settings, Delete account: that permanently removes your account, plan, sessions, sets, meals, photos, food snaps, habits, Apple Health numbers, crew posts, comments and photos, and Stoke conversations. Backups roll off within 30 days. To take back a permission without deleting anything, use Settings: Stoke and AI, Stoke reads Health numbers, or iOS Settings for Apple Health, the camera, the microphone and notifications. You can also email us to see, correct or delete what we hold.

Not medical advice

Almost Fire gives training guidance, not medical advice. It never decides you’re medically cleared to train. Check with a doctor before starting a new program, and if anything hurts. See the terms.

Children

Almost Fire is for people 16 and over. We don’t knowingly collect data from anyone younger.

Changes

If this policy changes in a way that matters, we’ll tell you in the app before it applies.