Parallax

Sign in

Username Password

First run: create the owner account

Run parallaxctl auth bootstrap on the server and paste the one-time token.

Bootstrap token Your name Username Password (12+ characters)

Allow listening on this browser?

When you start listening, this tab uses your microphone. It detects speech locally and sends only speech segments to your Parallax server, which transcribes them and decides whether you were talking to the assistant.

Audio is kept in memory only (at most 30 seconds) and is never stored.
Transcripts are not stored unless you choose a retention period in Settings.
The assistant only answers when addressed by name. Overheard needs ("we need milk") become silent suggestions only for categories you enable.
Nothing is sent, bought or changed without your approval.
Pause or close the tab at any time; revoking consent stops capture immediately.

Consent text version . This browser does not identify who is speaking: everything is attributed to you, the signed-in owner of this device.

keep listening while this tab is hidden

Say “Jeff, …” to talk to the assistant. Everything else is treated as background conversation.

Suggestions from background conversation, memory candidates and drafts. Approving records your decision: no message is sent, nothing is bought, no device is changed (no integrations are connected).

Assistant name(s), comma-separated

Passive capture from background speech (off by default) shopping needs (“we need milk”) reminders (“I should call grandma”) appointments likes and dislikes (“I liked that”) personal facts (“I'm allergic to…”) purchase ideas (draft only)

When I ask directly, add list items and reminders right away (undo in inbox) Speak answers aloud (only when I address the assistant) Follow-up window (seconds) Keep transcripts for Language

What the assistant decided for each utterance. No audio is stored; transcripts appear only within your chosen retention.