ShogunAI reads the text macOS already exposes to screen readers, keeps it encrypted on your Mac with its source and time, and indexes it on-device.
The useful part of a workday is never in one document. It is spread across a thread you skimmed, a decision you made out loud, a draft you abandoned, and a document someone else edited. ShogunAI keeps that as a memory you did not have to maintain — and keeps it where you can see exactly what it holds.
Text first, through the accessibility layer
The focused window’s app, title and visible text, read the way a screen reader reads it. Excluded apps, private windows and password fields are never read, and pause stops capture entirely — those exclusions apply before anything is written, not as a filter over stored data.
Screenshots where text falls short
Visual recall is off until you turn it on. With it on, a compressed screenshot is kept when a window yields no text or very little — under 100 characters, or 400 during a transcribed meeting — when it is a canvas-style app such as a design board, and, during a meeting, when it is a presentation or a shared screen: a chart in a shared slide, a canvas dashboard, a scanned PDF. Excluded apps are never captured. It stays encrypted on the Mac and deletes itself at the retention age you choose, one to seven days or a bounded custom duration.
Indexed on your Mac
The shipped app uses a local lexical text index and does not bundle or download an embedding model. Text recall works without a network, and your memory is not sent to a cloud embedding service.
Every memory carries its source
Entries link back to the event that produced them. That is what lets recall answer "Tuesday 14:20, in the design doc" instead of a confident guess, and it is what keeps low-confidence guesses out of anything you send.
From context to a useful next step
The useful part of a workday is never in one document. It is spread across a thread you skimmed, a decision you made out loud, a draft you abandoned, and a document someone else edited. ShogunAI keeps that as a memory you did not have to maintain — and keeps it where you can see exactly what it holds.
Grant accessibility once
macOS asks, you approve, and capture follows the window you are actually working in. Nothing is read from apps you excluded.
Work as usual
Context accumulates without note-taking. Tiered storage keeps the last day instantly available, the last month in full detail, and older history compressed.
Inspect and prune
Delete an entry and it leaves search, storage and any context assembled for a later action. Pause whenever you want; captured nothing means nothing stored.
What this changes
Spend less time rebuilding context
Recover the reasoning behind a decision weeks later
Find the source of a number instead of re-deriving it
Return to interrupted work without rebuilding the story
Keep sensitive context on the device by default
Frequently asked questions
Clear answers before you start
Does ShogunAI take screenshots?
Only if you switch visual recall on — it is off by default. Then a screenshot is kept where text capture returns nothing or very little (under 100 characters, or 400 during a transcribed meeting), for canvas-style apps such as design boards, and for presentations and shared screens during a meeting. Excluded apps are never captured. Those screenshots stay encrypted on your Mac, can be viewed or deleted at any time, and are deleted automatically at the retention age you pick. Everywhere else, capture is text through the accessibility layer and no image is written.
Is meeting audio stored?
No. Audio streams through your Deepgram account for live transcription, with model-improvement opt-out enabled, and ShogunAI never writes it to disk or a temp file. What persists is the transcript text and where it came from.
Where does memory live?
In an encrypted database on your Mac. There is no server-side copy of your day to seize, breach, or quietly train on. What leaves the device is the specific slice a request needs, logged with its purpose.
Does a heavy day cost more?
There is no per-token cloud fee for text indexing. The lexical index and text recall run locally and work offline.
Explore all Features
See how ShogunAI connects private memory, recall, and execution.