Shadow

Shadow is a Mac AI meeting assistant from Taper Labs, Inc. that captures what you say, hear, and see during calls, then runs your own prompts — called Skills — on that context. It's aimed at Mac professionals who live in calls and text fields: consultants, sales reps, recruiters, PMs, founders, and diligence teams who want meeting notes ready the moment a call ends and fast drafts without switching apps. The core workflow splits in two. Autopilot Mode detects a meeting, records on its own, transcribes with real-time speaker identification, and captures the meeting window (Zoom, Huddle, anything) as a Smart Screenshot — Shadow never joins the call as a bot. When the meeting ends, Shadow runs a Meeting Skill you pick or write: meeting notes, a BANT breakdown, action items, a follow-up email, even a source-linked incident postmortem or a diligence evidence ledger. Because screenshots are captured alongside audio, the Ask feature can query what was shown, not only what was said. Outside calls, Action Skills fire from a keyboard shortcut in any app. Shadow grabs voice input, the active window, or selected text, runs your prompt, and pastes the result at your cursor or into the focused field. Each Skill can carry its own dedicated shortcut, and Quick Access launches or searches Skills from anywhere. Transcripts and outputs export as clean Markdown and can push to Zapier, n8n, or Make via webhook. Privacy is the design constraint rather than a marketing page: transcription runs on your Mac, audio never leaves the device, and Skills only reach LLM APIs when you trigger them. AI can be turned off entirely, and meeting content is never used for model training. Recent vendor posts push the same angle from a permissions standpoint — a 2026 post on Apple tightening Full Disk Access lays out what a Mac AI assistant should be allowed to request. Against Otter, Fireflies, and Fathom, the edge is being bot-free, screen-aware, and Mac-native.