Last updated 30 September 2026.
Loomi is a personal assistant operated by Saif Alhameli as a sole proprietor, currently running as a closed pilot for a small number of invited testers. This policy describes what Loomi does with your information. Contact: saif.salem.alhameli@gmail.com.
Account details you give us: your email address, your name as Google reports it, your mobile number if you choose to add one, and your time zone.
Access to accounts you connect: if you connect Google, we store an encrypted token that lets Loomi read your mail and calendar and create drafts and events. If you connect Canvas, we store your Canvas access token, encrypted. If you connect Granola, Loomi reads your own meeting notes, read-only, to capture follow-ups and prepare you for meetings. Tokens are encrypted, you can disconnect any time in Settings, and they're deleted with your account. Granola is a processor, the same as Google and Canvas. You can disconnect any of these at any time.
What Loomi works out about you: commitments it believes you have made or are owed, the people those involve, notes about them, and your upcoming coursework. This is derived from your mail, calendar and Canvas.
A record of what Loomi did: every tool it used, every draft it created and every message it sent, so there is an auditable history. Long values are truncated and anything that looks like a credential is redacted before it is written.
Loomi does not sell your information, and does not share it for anyone else's advertising or marketing.
We do not share, sell, or provide your mobile phone number or messaging consent data to third parties or affiliates for marketing or promotional purposes. Your number and your consent are used only to send you the messages described in the Terms.
Loomi sends roughly one to five messages a day, depending on what is happening in your week. Message and data rates may apply. Reply STOP to stop, HELP for help.
Loomi cannot send email as you. It creates drafts in your own mailbox and you press send.
Loomi does not store copies of your mailbox. Messages are read when a task needs them and are not retained wholesale; what persists is the small set of facts described above.
Loomi runs on Vercel, and its database is hosted by Neon. Text messages are sent through Twilio. Your connected accounts are Google and, if you add them, Canvas and Granola.
To understand your mail and answer your questions, Loomi sends relevant text to language and evaluation models through the Vercel AI Gateway. That text can include the content of your messages. Those providers process it under their own terms. The evaluation model used for routine judgements is configured for zero data retention; we have not yet obtained written confirmation of retention terms for every model in use, and we will update this page when we have.
One of those evaluators is TypeSafe AI's Jev model, which we also use for the decision log described below. Our other processors are listed above: Vercel, Neon, Twilio, Google and, if you add them, Canvas and Granola.
We do not otherwise disclose your information, except where we are legally required to.
Loomi keeps a decision log of the small yes/no calls it makes as it runs — whether to notify you, whether to read an email, which item leads a brief — so it can get faster, cheaper and better at knowing when to speak up. Each record holds: the kind of decision, the evaluator's confidence, the threshold it was checked against, whether Loomi acted, what you did next (the item got done, was dismissed, muted or ignored), and content-free context such as the item type, hours until due, whether the counterparty is a known contact, and the local hour.
A record never holds the words of a message or email, a subject line, a name, an address or a phone number. Items are identified by a one-way hash, not by their content, so a record cannot be turned back into the message it came from.
Records expire automatically after 180 days, and are deleted immediately if you delete your account.
This is on by default, with this disclosure, and you can turn it off at any time in Settings or by texting IMPROVE OFF. While it is off, Loomi records none of these decisions for you; text IMPROVE ON to turn it back on.
We use this only to tune Loomi's own thresholds and to train Loomi's own small decision models. We do not sell it, and we do not use it to train any other company's models.
The decisions themselves are made by TypeSafe AI's Jev model, one of the evaluators named above. To decide, it sees the details of the item in question (for example an email's sender and subject), the same way the language model providers already described do, under their own terms. What we keep afterwards is only the content-free record above.
Loomi keeps the last two weeks of texts between you and Loomi, in both directions, so it can follow the conversation even when something else — a scheduled brief, a meeting reminder, a fresh check from Loomi — starts a new session partway through. These are stored encrypted, the same way as everything else worth protecting (see Keeping it safe below).
Only your own Loomi reads them. They are deleted automatically after 14 days, and immediately if you delete your account. They are never used to train models — ours or anyone else's — and never shared.
Access tokens are encrypted before they are stored, with a key held outside the database. Sessions are signed. Only you can see your own items: every read is scoped to the authenticated account. The assistant has no shell, no filesystem access and no ability to browse the web.
No system is perfect, and Loomi is an early pilot rather than a mature product. You should connect it knowing that.
You can download everything Loomi holds about you, and you can delete your account and its data, from your account page at any time. Deleting your account does not revoke Google's grant — do that at myaccount.google.com/permissions.
If you join the waitlist, we keep the details you enter only to contact you about access, and delete them on request.
When an account is deleted, we keep anonymous cost records — which service was used, how much it cost, and when — so our running totals stay correct. They contain nothing you wrote and nothing that identifies you.
If you have texted STOP, we keep a record that your number should not be texted again, even after account deletion, so that a number is never silently re-enrolled.
You can stop messages at any time by replying STOP, and get help by replying HELP.
If this policy changes in a way that affects what Loomi does with your information, we will tell the people it applies to. The date at the top reflects the current version.