Everything we promise you, in writing.

Your Twitch login and user id, used to run your account. Nothing more is needed to sign you in. We also read the email address on your Twitch account when you sign in, and keep it encrypted, so we can send you account emails: a new sign-in, a change to your access, notice before our terms change, notices about your account itself, such as a security incident, and, if we invite you to be a partner, the invitation and anything that changes about your partnership. It also carries any notifications you choose to get by email, and marketing only if you opt in. If you check a referral code before subscribing, that code is another streamer's Twitch username and is sent so we can tell you whether it is valid.

Channel point redemption events for your own Pro analytics: the reward, its cost, and who redeemed. Each viewer's id is hashed to a per-channel pseudonym so it cannot be linked across channels, and their display name is stored encrypted at rest so your analytics can show your own top redeemers by name. Used only to build your analytics, never to profile viewers across channels. Pro only.

The category, subject, and message of any support ticket you open. Subject and message are encrypted at rest.

Direct messages are end-to-end encrypted; the server only ever holds ciphertext it cannot read.

Whether your Pro subscription is active and when it renews, and, if you subscribe using somebody else's referral code, the link between your account and theirs, kept so the reward is applied once. Payments are handled by Stripe; we never see your card details.

The content of anything you post to the Community Hub: feature requests, forum posts and replies, and bug reports, together with your Twitch login. This is shown publicly under your username where the Community Hub itself is public (the forum and feature requests, for example); it is not held privately like support tickets or messages. Survey answers are the exception and work the other way round: they are stored without your login, so we can see that you answered and never which answers were yours, and they are never shown publicly. The record that you took part, which is what pays the reward, is kept separately under your login.

If you use the mobile approval queue or the companion web app, a redemption waiting for your decision (the reward, its cost, and the viewer's name and typed message, where applicable) is held briefly on the server so those surfaces can show it to you. It is automatically deleted after 4 hours regardless of whether you acted on it.

If you report a forum post or reply, we hold your report: your Twitch login, which post you reported, the reason you chose, and any details you add (encrypted at rest). Reports are private and visible only to the moderation team; they are never shown publicly and the person you reported is not told who reported them.

The devices signed in to your account: a coarse device name (like SparkStream Desktop on Windows) and when each device first signed in and was last active. This powers the Active devices list in Settings and the email alert when your account signs in on a new device. No IP addresses, precise locations, or hardware identifiers are collected.

Which notification channels (in-app, desktop, email) you have turned on for each category, saved against your Twitch login. This is stored so email notifications honour your choice and so your preferences follow you across devices. It records your settings only, never the content of any notification.

If you turn on Cloud Sync, copies of your app configuration (your rewards, triggers, overlays, and settings, never your Twitch credentials or keys) is encrypted on your device with a passphrase only you know and stored so your other devices can pull it. The server only ever holds ciphertext it cannot read, plus the non-secret salt needed to derive the key from your passphrase. If you turn on restore points, we keep the recent versions rather than only the newest, so you can put your setup back to how it was on a particular day. How many are kept and for how long is shown in the app before you turn it on. Every version is ciphertext under your own key exactly as the newest one is, so keeping more of them tells us no more than keeping one did. This is entirely opt-in and off unless you enable it.

If you add the SparkStream Discord bot to your server, we store the link between that Discord server and your account: the Discord server's id, your chosen announcement channel and optional ping role, your go-live message, and the social links you set for the bot's /socials command. This lets the bot post when you go live and answer commands. It is opt-in (only if you add the bot), holds ids and your chosen settings rather than any message content, and is removed when you unlink the server.

If you turn on a Discord activity feed, SparkStream sends the matching live events to our server so the bot can post them to your chosen Discord channel. Depending on which feeds you enable, this can include: for the redemption feed, a channel-point reward's name, the viewer's display name, the cost, and any message they typed with it; for the now-playing feed, the title, artist, and link of the track you are currently playing; for the leaderboard feed, the ranked viewer display names and their totals (such as top gifters or cheerers) for the ranking and time span you choose; for the milestones feed, the name of a completed community goal, for an incoming raid the raider's display name and viewer count, and the follower or subscriber count when you pass a milestone; for the clips feed, a clip’s title, its link, and the reward or moment that caused it to be made. These are passed straight through to Discord and not stored on our server (the current track and the latest leaderboard are held only briefly in memory so the /song and /leaderboard commands can answer). Every feed is opt-in and only the feeds you turn on are sent.

If you add SparkStream's hosted Twitch chat bot to your channel, we store the link between your channel and the bot (your Twitch login and user id) along with the settings you choose for it: your custom commands and their responses, the command prefix, which built-in commands are on, any timers and the messages they post, any keyword replies and the phrases that set them off, and your chat moderation settings including any words or web addresses you add to them. So that it can answer, the bot reads your channel's chat to spot a command, one of your keyword phrases, or something your moderation settings tell it to act on. If you set rewards up to be redeemed while you are away, we also hold which of those rewards the bot should acknowledge and the message it should reply with, so that somebody redeeming one is answered at the time rather than the next time you open the app. It does not store your chat. The one exception is moderation: with it switched on, every action it takes is recorded so that you can see what happened and undo it, which means who it acted on (their Twitch name and id), what was matched, and a short extract of the message. Those records are kept for 14 days and then deleted. It is opt-in (only if you add the bot) and everything is removed when you remove the bot from your channel. These are the bounds: settings you typed, for a bot you added, used only to answer in your own chat. Anything the bot keeps about a viewer is the moderation record described above and its fourteen days.

If you start or join a collab session with another SparkStream streamer, we relay that session between the two of you: which accounts are linked, and, while it is running, the figures and events each of you has chosen to share. Depending on what you switch on, that can include a shared goal's progress, a shared countdown, combined leaderboard standings, and the fact that a reward was redeemed on the other channel together with its name and cost. Your viewers' display names are included only if you turn that on, and it is off by default. All of it is passed straight between the two apps so the other streamer's overlays can show it, and none of it is kept on our server once the session ends. Channel points never cross channels: a redemption is always settled on the channel it happened on. A session is opt-in, both streamers have to accept it, and it ends when either of you ends it.

SparkStream can record how much memory and processor it is using over time, so that a problem which only appears after several hours has something to show for it. That recording is written to a file on your own computer and stays there. If you decide to send it to us with a support ticket, we receive the recording and a short description of your computer: the processor model, how many cores it has, how much memory is installed, your operating system version, and your SparkStream version. We ask for those because the same numbers mean very different things on different machines. Nothing about your stream, your channel, your chat or your viewers is in any of it. None of this is ever sent on its own: starting a recording is something you do, sending one is a separate decision you make afterwards, and if you never send one then we never learn it existed.

If you turn on your public channel page, SparkStream sends the parts of your channel's state you have chosen to show there to our server, so that the page can show them to anyone who visits it. Depending on which panels you turn on, that can include: your channel point rewards and their current prices, the progress of your community goals, a running subathon clock, how many redemptions are waiting in your queue, a running auction's current bid, how long is left on it and how many people have bid, your custom chat commands and their responses, the quotes you have saved, and a short description of your channel if you choose to write one. That description is the one thing on the page you write yourself rather than something the app already had, so it is checked for slurs and hate before it goes up, and anyone who visits your page can report it. Whether you are live, your title, your game, your upcoming schedule and your profile picture come from Twitch rather than from your computer. Anything concerning your viewers appears as totals only: no viewer's name, id or message is published on the page, and an auction shows how many people have bid rather than who they are. The links on the page are a platform and a handle you choose from a fixed list, never a web address typed in free text, and no images are ever uploaded. The page is off until you switch it on, every panel on it is one you chose to show, and switching it off stops the page being served and removes what was held for it.

Some of the bot's features need to know what is happening in your app before they can answer. Each one is off until you switch it on, apart from the command list described below, and switching one on sends us only what that feature needs, and only while it is on. The subathon command is the first: with it on, we hold your subathon's current state, which is whether one is running, the time left, your rates and your multiplier, so the bot can answer when somebody in your chat asks. The !commands list has no switch of its own, because it is part of having the bot at all: while the bot is in your channel, we hold the names of the chat commands your app answers and when your app last sent them, so the list is complete and the bot can tell a viewer when one of those commands is not answering. Only the names, never what the commands say, and they go when you remove the bot. Some features will need to hold something about the viewers who use them, because a giveaway cannot draw a winner without knowing who entered. Where that happens it is the least the feature can work with, and it goes when you switch the feature off or when the thing it belongs to ends. Reading your chat to spot a command still does not mean storing your chat. All of it is used to run the feature you switched on and nothing else: not to build a picture of you or your viewers, not sold, not used to train anything. If we ever want to use it in a way this does not describe, that is a new category in this charter, and you hear about it before it happens rather than after.

While you are live, SparkStream reads your own channel's numbers about once a minute and keeps a running summary of that stream. It is the summary we hold rather than every individual reading: your average and highest viewer count, how long you were live, your subscriber and follower totals at the start and at the end, how many people followed and subscribed while you were live, whether the app was listening for the whole of that stream, and the title and category you streamed under. These are your own figures, the ones Twitch already shows you about your own channel, and they are totals. Nothing here describes an individual viewer, and no viewer's name, id or message is any part of it. It is stored encrypted at rest, and deleted once it is older than your analytics can reach back. Your Pro analytics is the first thing built on it. If we later build something else on these same figures we will not send you another notice, because it is the same numbers used in the same way, and these are the bounds that make that safe: your own channel's totals only, gathered only while you are live, used to run features you can see and nothing else, never sold and never used to train anything. Anything wider, and in particular anything about an individual viewer, is a new category in this charter, and you hear about it before it happens rather than after. Gathered only while your Pro subscription is active, and it stops when that does.

If you run SparkStream on two computers, a gaming PC and a streaming PC, Two-PC Mode passes what the two need to agree on between them through our server, because they cannot see each other directly. That is state your app already holds: which of your rewards are on, what your overlays are showing, what is waiting in your queue, and which machine is currently in charge. It is passed straight between your own two machines and is not kept once both have disconnected. Nothing about an individual viewer travels this way, and anything concerning your viewers is a total, in the same shape your overlays already show it. Both machines are signed in as you, and pairing them is something you do deliberately. These are the bounds: your own state, between your own machines, held only while they are connected. Anything about a viewer as an individual, or any use of this beyond keeping your two machines in step, is a new category in this charter and you hear about it first.

Your overlays normally run on your own computer and OBS reads them from there, which is why they keep working when your internet does not. If you switch an overlay to a hosted address instead, so it can be read from another machine, SparkStream sends that overlay’s current state to our server so the page can be drawn. Depending on the overlay, that can include a goal’s progress, a subathon clock, an auction’s current bid and how long is left, a wheel’s segments and where it landed, leaderboard standings, and how many redemptions are waiting. Anything concerning your viewers appears as totals, or as the display names you had already chosen to show on screen, and nothing is added that your overlay was not already showing. A hosted overlay is off until you switch it on, it is served only at the address you are given, and switching it off stops it being served and removes what was held for it. These are the bounds: the state one of your own overlays is already displaying, held only while that overlay is hosted. Anything wider is a new category in this charter.

Engagement analytics answers which of your rewards actually bring people into your chat, rather than which are simply redeemed most. Working that out means looking at your own channel’s redemptions and your own channel’s chat together, viewer by viewer, and that work happens on your computer rather than on ours. If you turn on analytics history, the result is encrypted on your device with a key only you hold and stored so it survives a reinstall and reaches your other machines. The server holds ciphertext it cannot read, plus the non-secret salt needed to derive your key from your passphrase. We cannot see which viewer did what, we cannot compute anything from it, and we could not answer a question about it if we were asked. Storing your history is opt-in and off unless you turn it on, and turning it off deletes what was held. These are the bounds: your own channel, analysed on your own machine, stored only as ciphertext we cannot read. If anything here ever needed to be readable by us, that is a new category in this charter and you hear about it before it happens rather than after.

If you are a SparkStream partner and choose to write a testimonial, SparkStream sends the words you write to our server, together with your Twitch login. A member of our team reads it, and if they approve it, it is shown on the SparkStream website with your channel name and a label saying you are a SparkStream partner. It is shown exactly as you wrote it: we can decline to put it up and tell you why, but nothing in SparkStream can change what you wrote. It is checked for slurs and hate before anyone reads it. Every version you send is kept with what happened to it, so you can see what was shown. You can take it down at any time, and it is deleted when your partnership ends or you delete your account. Nobody who is not a partner can send one.

Once a day, a list of true or false, one entry per feature, saying whether each feature is switched on. It is used for one thing: to see what people actually use, so we know what to build and improve next, and which features to go and ask people about rather than quietly leave to rot. That is the whole of it. Not how often you use anything, not when, not for how long, not in what order, not what you set it to, not whether it worked, and never the content of anything. It is counted the moment it arrives and then discarded: the server keeps a count for each feature, and separately a note that your account reported that day, never the two together, and the note is deleted when the day closes. So no history of you is kept, no combination of features is ever recorded, and what is left is a number with nothing attached to it. Your list itself is never written down, so there is nothing of it to encrypt or to hand over: it travels over an encrypted connection and exists only while it is being counted. Nobody at SparkStream is shown who uses what, because nothing holds that. The only thing anyone here sees is the totals, and a total under five is shown only as fewer than five. It is never sold, never used for advertising, and never used to train anything. This is the one category that is on unless you switch it off, because it is not tied to a feature you turned on. You can switch it off at any time in Settings under Backup & Data, where you can also see the exact list before it is sent, and if this reached you as a change to the charter, on the notice that tells you about it. Switching it off stops every future report. A report already counted cannot be taken back, and deleting your account does not reduce a count, because nothing records which number was yours. These are the bounds, and they are a promise about what this must never become: one true or false per feature and nothing beside it. Counting how often or in what order, what anything is set to, whether it worked, keeping anything about you past the day it was sent, or breaking the totals down by plan, region, how long somebody has been signed up or anything else about the people counted, is a new category in this charter, and you hear about it before it happens rather than after. There is no exception for one person or one investigation, because there is nothing to look in.

If you link an account on another streaming platform so that its chat can appear in SparkStream, we store the link between that account and yours: the platform, your user id and username there, and the request that asks that platform to send us your channel's chat. Kick is the first such platform. Kick only sends chat to a web address, so each message reaches our server and is passed straight on to your app: the chatter's username on that platform, the colour and badges it shows beside their name, the message itself, and which message it replies to, if any. Messages are not kept. Each one is on our server only for as long as it takes to pass it on, and it is used for nothing else. Your sign-in to that platform stays on your computer; it is used once, when you link, to show us the account is yours, and we do not keep it. It is opt-in, only if you link an account, and when you unlink it we delete the link and ask the platform to stop sending. If you choose to show that chat in one of your overlays, it appears on your stream as it appears on the platform it came from. These are the bounds: your own channel's chat, on a platform you linked, passed to your own app and kept nowhere. Another platform is named here before it is offered. Keeping any of this chat, or using it for anything but showing it to you, is a new category in this charter, and you hear about it first.