MeetVoiceOS.Say it and it’s done.

• Use dictation mode to turn speech into text and agent mode to request actions rather than only transcription.
• Point at screen context and ask the assistant to locate relevant files, emails, documents, notes, or web results.
• Use supported connections and applications for actions such as drafting or sending email, replying in Slack, scheduling meetings, creating notes, and filing tickets.
• Keep a requested task active so VoiceOS can bring it back at an appropriate time according to its current agent workflow.
• Control whether transcripts are saved on the device and whether anonymous diagnostic or crash reports are shared.
• Rely on the developer’s stated audio policy: audio is not stored on its servers and is not used to train models.

VoiceOS is for people who want voice input to start or complete multi-app work, not merely dictate text. It may be useful for repetitive communication and retrieval work when the available integrations match the tools you already use.

Treat every external action as something to review before it is sent or committed. It is not a substitute for checking facts, permissions, recipients, or the security of the accounts and files connected to it.

VoiceOS needs microphone access and can request screen, file, application, and connected-account context to complete actions. The developer says audio is not stored on its servers or used for training, while transcripts are saved locally unless the user chooses otherwise. Sending mail, messages, tickets, or calendar changes remains an external action that should be reviewed before execution.

VoiceOS is subscription-based: $11.99/month billed annually.