What does Littlebird actually see?
Littlebird reads the text and elements of the active window you are working in, via macOS accessibility permissions you grant, and during meetings it listens along to transcribe and summarize, with no bot joining the call. You can pause capture, exclude apps, and delete data at any time. That is the whole mechanism, stated as precisely as we can verify it.
What "the active window" means
The window you are actually working in: the doc you are writing, the site you are reading, the thread you are answering. The app pays attention to that, which is why the shorthand for the mechanism is screenreading, reading the text on your screen. It is how the assistant already knows your work when you ask about it later, without you filing or forwarding anything.
How meetings get captured
During meetings the app listens along on your computer to transcribe and summarize, and no bot joins the call. The same mechanism covers any audio or video playing on your computer, like a recorded webinar, not only live calls. One limitation stated plainly: it does not currently claim accurate speaker attribution, so if you need reliable per-speaker labels, test that against your own calls first.
The permissions it needs, and what you control
On Mac, the app requires user-granted macOS accessibility permissions; nothing is collected without your permission. From there the controls are yours: decide whether it runs at startup, pause context collection whenever you want, exclude specific apps (your bank, your journal, a client's privileged material), and delete your data at any time, all of it or just the last hour or day. Deleting your account deletes the memory with it.
Where what it sees ends up
In an encrypted memory stored in the AWS cloud (US East). AES-256 at rest and in transit, TLS 1.3 connections, SOC 2 certified, regularly audited by third-party security firms, never used to train models, never sold. The full picture, including who should not use a cloud-based tool at all, is in Is Littlebird safe?, and the formal documentation lives at trust.littlebird.ai.
What we don't claim
A section most vendors skip. There are capture questions people reasonably ask, like exactly how the app handles passwords, private windows, or visual content, where the precise engineering answer has not been through our verification process. Until it has, we do not publish claims about them, in either direction. What we publish is the verified mechanism above and the controls you can test yourself: turn on app exclusion, pause capture, and watch the behavior.
Common questions
How does Littlebird understand my work?
It pays attention to the active window (the doc, the site), listens during meetings, and can optionally connect email, calendar, and apps.
Can I stop it from seeing something?
Yes. Pause context collection, exclude specific apps, or delete data from the last hour, the last day, or all time. You decide if it runs at startup.
Does a bot join my meetings?
No. It listens along on your computer to transcribe and summarize, and no bot joins the call.
Where does Littlebird store my data?
In an encrypted memory in the AWS cloud, US East region. AES-256 encryption at rest and in transit, TLS 1.3 on every connection, SOC 2 certified.
Does Littlebird train AI models on my data?
No. Your data is never used to train models and never sold. It is used only in ways you explicitly approve.
Consent-based capture, with the controls in your hands.
Install it, grant exactly the permissions you choose, and test the exclusions yourself.
Capture-mechanism language on this page uses only phrasing verified in our maintained internal fact base.
