Privacy Policy

Last updated: November 10, 2026

Atlas is a personal AI assistant you text over iMessage (delivered through Photon) and use in the Atlas iOS app. This policy explains what Atlas collects, why, who it is shared with, and the choices you have.

Who is responsible. Atlas is operated by Hypurr Inc. ("we", "us"), the controller of your personal data. Contact: hi@hypurrclaw.xyz.

The short version

We use what you send Atlas to answer you and do the things you ask.
To answer, your messages are sent to the AI model provider powering your Atlas (see "How the AI works").
We do not sell your data, show ads, or track you across apps and websites.
We do not use your content to train our own models.
You can export your data or delete your account at any time from the Atlas app.

What we collect

Messages. What you send Atlas and Atlas's replies, over iMessage and in the Atlas app.
Photos and files you send or upload.
Voice. Voice recordings you send and their transcripts.
Memory. What Atlas learns about you to be useful over time: a core profile (for example your name, preferences and standing instructions) and a graph of people, places, things and facts mentioned in your conversations, with when each fact was true.
Automations and reminders you set up, and their run history.
Connected services. If you connect a service such as Gmail, Slack, Asana or an MCP server, the access tokens that service issues so Atlas can act for you, and the data Atlas reads or sends there at your request.
Saved observations and procedures. Bounded copies of text and images Atlas reads while doing your work, with source and time metadata so you can check citations; reusable procedure versions and their execution and review history; and browser action status records used to avoid repeating an interrupted action.
ChatGPT or Grok login. If you choose to power Atlas with your own ChatGPT or Grok plan, the login tokens OpenAI or xAI issue for it.
Identity. Your phone number or iMessage handle, which is how your account is identified, and the name you give Atlas.
Device. Your Expo push token (to send notifications to the Atlas app), and your device platform and app version.
Notifications. With notifications on, the text of Atlas's replies (up to 1,000 characters) is sent through Expo to Apple (iPhone) or Google (Android) so your phone can show it, and a notification Expo couldn't take is kept on our side for about 45 minutes while we retry it. Turn off "Show Previews" in Atlas under Settings (next to Notifications) and a notification only says Atlas replied: the reply's text then goes to none of them.
Usage and diagnostics. Counts such as messages and automation runs (used for free-tier limits and the usage screen), and what a problem report carries: your description, app version and build, platform, the last error the app saw, and a pointer to your most recent chat messages so we can find the conversation you're reporting about. When the app crashes, it sends a crash report the next time it opens: the error and where in the app's code it happened, the app version, build and update, and your device's platform and OS version. It's kept with your problem reports. Crashes in the app's native code also go to Google Firebase Crashlytics with your device model, OS version, app version and an ID for the app installation; they carry no messages or files.
Messaging preferences and fair-use counters. For each number or handle that texts Atlas: whether it texted STOP or START, whether it has seen the notice that Atlas is an AI, and recent message counts used to apply fair-use limits.

We do not ask for your contacts, location or advertising identifier.

Why we use it

To provide Atlas: receive your messages, answer them, and carry out what you ask, such as searching, drafting, scheduling or sending on your behalf.
To remember context, so you don't have to repeat yourself.
To run your automations and reminders and deliver notifications.
To enforce free-tier and fair-use limits, honor STOP requests, prevent abuse and keep the service secure.
To give support and fix problems you report.

Legal bases (EEA and UK users)

Contract: processing needed to provide the service you asked for, including sending your content to the model provider and to services you connect.
Legitimate interests: keeping Atlas secure, preventing abuse, enforcing usage limits and fixing problems, balanced against your rights.
Consent: where we ask for it, such as push notifications, connecting a service, or a ChatGPT or Grok login. You can withdraw consent at any time by turning the feature off or disconnecting; this does not affect processing that already happened.

How the AI works

Atlas is powered by large language models run by other companies. To answer you, your message and the context Atlas needs (recent conversation, relevant memory, files you shared and results from tools) are sent to the model provider powering your Atlas:

Free (default): Anthropic, under our account. When it's unavailable to us, OpenCode (Go) routes requests to the model vendor behind its free model, such as Anthropic, OpenAI or others.
ChatGPT: OpenAI, under your own ChatGPT plan and account.
Grok: xAI, under your own Grok plan and account.

Speech-to-text, text-to-speech and web search run on Cloudflare Workers AI. We do not use your content to train our own models. How a model vendor handles the data it receives is governed by its own terms; for ChatGPT and Grok, your own account's data settings with OpenAI or xAI apply. See Subprocessors for the full list.

Sharing

We share personal data only with the service providers listed on the Subprocessors page, with services you choose to connect, when you ask Atlas to send something to someone, or when required by law. We do not sell or share your personal information for cross-context behavioral advertising, as those terms are defined in the California Consumer Privacy Act, and we have not done so in the past 12 months.

How long we keep it

Your account data (memory, automations, settings, messaging preferences) is kept while your account is active. Chat messages are kept for two years: older ones are removed from your chat history and from the search over past messages, while what Atlas remembered from them (memory) stays until you delete it. Download your data from the app first if you want a copy of older messages.
Uploaded files are kept until you delete them or delete your account.
Saved observation bodies expire after 30 days. Procedure execution history can contain copies of tool results and remains with your account until you delete a settled run or the account. Deleting a run retains content-free execution identities and outcome counts to prevent accidental replay and duplicate credit.
Problem reports are kept with your account (the latest 50) and deleted with it.
Backups of account storage are rotated out within 30 days of deletion.
After you delete your account, we keep a minimal record that it was deleted, with no message content, and your phone number or iMessage handle can't set up a new account for a short period while deletion completes.
Operational logs are short-lived and never include message text.

Deleting your data

In the Atlas app, go to Settings › Account › Delete account. This deletes your chat history, memory, uploads, saved observations, procedures and their history, browser action records, automations and connected-service links. Deleting also revokes the ChatGPT, Grok and Google logins at their providers, asks each other connected service to revoke the access it gave Atlas where that service offers a way to, and removes your number from Photon's list of users. A service that offers no way to revoke keeps the access until it expires, so we recommend also revoking Atlas's access there. If you can't use the app, email us and we'll delete your account after confirming it's yours.

Exporting your data

In the Atlas app, go to Settings › Account › Export my data to download a JSON file with your profile, chat, memory, automations, reminders, integrations, uploads (with download links), saved-observation metadata (with account-authenticated body links), procedure versions and execution history, browser action records and problem reports. Download observation bodies before their expiry; the export does not extend their retention.

Your rights

Depending on where you live, you have the right to access, correct, delete and port your personal data, to object to or restrict certain processing, and to withdraw consent. You can use most of these directly in the app; for anything else, email hi@hypurrclaw.xyz. We will not discriminate against you for exercising these rights. If you are in the EEA or UK, you can also complain to your local data protection supervisory authority.

Children

Atlas is not for children under 13, or under 16 in EEA countries where that is the age of digital consent. If we learn that an account belongs to a child below these ages, we delete it. If you believe a child is using Atlas, contact us.

Security

Each user's data is kept in its own isolated storage on Cloudflare. Data is encrypted in transit and at rest by our hosting provider, access tokens for connected services are stored server-side and never shown to the model, and the app authenticates with per-device session tokens. No system is perfectly secure; tell us right away if you suspect a problem.

International transfers

Atlas runs on Cloudflare's global network, so your data may be processed in countries other than yours, including the United States. Where we transfer personal data from the EEA, UK or Switzerland, we rely on the European Commission's Standard Contractual Clauses (and the UK addendum) or another lawful transfer mechanism.

Atlas is an AI

You are talking to an AI, not a person. The first time you text Atlas, it tells you it is an AI assistant and that you can text STOP to stop messages. It can be wrong, incomplete or out of date. It is not medical, legal, financial or other professional advice; check anything important.

Stopping messages

Text STOP (or STOPALL, UNSUBSCRIBE, CANCEL, END, QUIT, OPT OUT or REVOKE) as a message on its own. Atlas replies once to confirm, then sends nothing more to you on that platform and ignores your messages there. Text START (or UNSTOP, RESUME or OPT IN) to turn messages back on. We keep your STOP or START choice so we can honor it.

Changes

We will update this policy as Atlas changes and revise the date above. For material changes we will tell you in the app or by message before they take effect.

Contact

Hypurr Inc.: hi@hypurrclaw.xyz