VoxAI brings live meeting transcripts to your Mac via AI copilot

VoxAI is a native macOS application that combines local audio recording, real-time transcription with speaker identification, and integration with AI tools over the MCP protocol. The core workflow is simple: the app records conversations (meetings, calls, in-person discussions, or voice chats), transcribes them on-device with live speaker labels, and simultaneously streams the transcript to an AI client (such as Claude Code, Codex, or Cursor) that can monitor the conversation and push advice, questions, risks, and action items into a side panel while the meeting is still underway.

The app operates in four modes: Meeting (multiple speakers with live names and AI notes), Call (separate channels for your mic and system audio with automatic labeling), 1:1 In-Person (for negotiations, interviews, and one-on-ones), and Voice Chat (speak to the AI from across the room and hear it respond aloud). Because VoxAI separates microphone and system audio on calls, speaker identity comes from the channel rather than guessing, eliminating the ambiguity of cloud-based diarization.

Privacy is the default. Recording, transcription, speaker labels, and local voice processing all run on-device. Cloud features are opt-in only, behind explicit switches. The app requires no account, no API key, and no open network port. Users retain full access to their archive—they can view, export, and search every recording forever, even if they do not purchase a license.

Setup takes three steps: install the .dmg, copy a connection prompt from VoxAI and paste it into an MCP client, and choose a conversation mode. The MCP server exposes thirteen tools (list_meetings, get_meeting, rename_meeting, get_live_transcript, start_listening, stop_listening, push_advice, get_dialog_input, post_dialog_reply, speak, stop_speaking, list_voices, and update_voice_config) so the connected AI client can browse the archive, rename recordings, listen live, and deliver advice. After the first ten free recordings, the app degrades gracefully: new recordings cap at ten minutes, live copilot advice pauses, but the history remains fully viewable and exportable. Unlock pricing is US$19.99 via the Mac App Store or US$19.90 for the direct .dmg version; both are one-time payments with no subscription renewal. The app requires macOS 14 or later and Apple Silicon.

VoxAI addresses a practical gap in meeting workflows: most transcription and note-taking tools capture the conversation after it ends, leaving participants to replay recordings or wait for a summary. By streaming the live transcript to an AI copilot via MCP (Model Context Protocol), the app lets the AI agent insert advice, fact-checks, and action items into a side panel while the conversation is still happening—a real-time collaboration rather than post-meeting analysis.

The product's positioning centers on privacy and simplicity. Local processing (recording, transcription, speaker labels) means sensitive business conversations never leave the Mac unless the user explicitly opts in to cloud features. The setup is minimal: drag the app to Applications, paste a connection prompt into your AI client, and start recording. This stands in contrast to cloud-based transcription services that require accounts, API keys, and ongoing management.

The pricing model—a one-time purchase rather than a subscription—removes recurring friction, though the ten-session trial with degraded features (ten-minute caps, paused live advice) creates a natural funnel to unlock. The distinction between Mac App Store and direct .dmg purchases (with different key systems) mirrors common indie macOS app distribution but adds some friction for users who want to switch between them.