// support_channel
Mortise runs entirely on your Mac, iPad, or iPhone, so most problems are about memory, model choice, or downloads rather than an account or a server outage. There is no account to recover and no status page to check — just write to us.
Last updated 11 November 2026
// reach_us
No account needed, and no sign-in — write from whatever address you like. We reply to every email, usually within two business days.
Or copy the address: support@mortiseapp.com
// which_app
Private, on-device AI for Mac, iPad, and iPhone, published by Exception To Policy (Seattle, Washington). It ships with one open-weight language model, can download others, and runs them on your device's own processor — no cloud inference, no account, and no server of ours in the path. This page supports that app.
Which models run well depends on your hardware. The model catalog lists download size and estimated peak memory for every model, grouped by the device it suits.
// from_inside_the_app
Each one opens a draft in your own mail client. You see everything before it is sent, and you choose whether to attach recent error details and your app and device information — attaching them makes a bug far easier to diagnose.
On any reply that was wrong, offensive, or strange.
When something visibly breaks.
In Settings › General, for feature requests.
// objectionable_output
Language models can produce output that is offensive, harmful, or plainly false, and Mortise does not filter what a model generates. If a model produces something objectionable, tell us: use Report this response on the reply itself, or email support@mortiseapp.com.
We read these promptly and reply, usually within two business days. Reports directly shape which models we recommend and which we drop from the catalog — a model that repeatedly generates objectionable output gets removed. Using the App to generate content that is illegal or that harms others is prohibited outright; see the Terms of Use.
// common_fixes
A few things resolve most reports.
Larger models need more memory than some devices have. In Settings › Models each entry shows its download size and estimated peak memory — try one marked as suitable for your device. LFM3 1.7B, built into the app, fits most devices.
Models are several gigabytes. Cancel and restart it from Settings › Models — downloads resume rather than starting over. You don't need a download to start: the built-in model is already there.
Both devices need Mortise open, Device Sync turned on, and the same network. If you declined local network access, allow it for Mortise in the system's Privacy settings, then pair again.
Open links is off by default. Turn it on in Settings; until then a link is ordinary text.
Smaller models have shorter limits and can stop mid-sentence. Ask for a shorter answer, or switch to a larger model.
That is a known property of language models rather than a fault in the App — see the Terms of Use. Reports still help us tune which models we recommend.
// what_to_include
Your app version and device — both filled in automatically if you tick "app and device info" — which model was selected, what you asked, what happened, and what you expected instead.
// support_privacy
We only ever see what you put in an email and send yourself. The App sends nothing to us, in the background or otherwise. Device Sync, if you turn it on, talks only to your own paired devices on your local network — see the Privacy Policy. If your conversation is sensitive, describe the problem instead of pasting the transcript.
// buy_once
No subscription, no account, no server of ours. Every update included.