Capy

Live captions, translation and recaps of any conversation on your Mac.

Capy is a macOS app that captions, translates and summarizes any conversation in real time: a meeting, a video, a call in your headphones. It runs on Zoom AI Services: get an API key there and try it in minutes.

Capy is an independent project, not affiliated with or endorsed by Zoom.

Powered by Zoom AI Services

Zoom AI Service	What Capy does with it
Scribe	Live captions over every app, from your microphone or anything your Mac plays, with speaker labels
Translator	Real-time translation of every caption, across 10 languages
Summarizer	A recap with action items for every session
Speech Summarization	Transcripts and recaps of audio and video files
Decisions	Highlights of what matters: requests for you, decisions, debates, action items

Audio is never stored. Transcripts stay on your Mac.

Requirements

• macOS 13+ (14.2+ to caption what the Mac plays)
• Go 1.27.1+
• A Zoom AI Services API key from zoom.ai

Quick start

make run        # run from source; a capybara appears in the menu bar
make run-fresh  # same, on an empty library, as at first launch
make bundle     # build/darwin-arm64/Capy.app and Capy-<version>.dmg

Then open Settings (gear in the sidebar, or Settings… in the menu) and paste your API key.

API key

1. Sign in at zoom.ai.
2. Create an API key with the permissions you need: Scribe, Translator, Summarizer (with Speech summarization for Import) and Decisions (for Highlights).
3. Copy the key; it is shown only once.

The key is stored in the macOS Keychain; $ZOOM_API_KEY overrides it. Other programs you run can read it from the Keychain without a prompt, so limit the key to the permissions above.

Permissions

macOS asks for Microphone on first use, and for System Audio Recording the first time Captions listens to "This Mac". Capy hears audio only, never the screen. Notifications work only in the bundled app.

Signing

make bundle signs ad hoc, so macOS asks for permissions again after every rebuild. To keep them, save your signing identity (gitignored):

security find-identity -v -p codesigning                     # list yours
echo 'Apple Development: Jane Doe (TEAMID)' > .signing-identity

entitlements.plist grants the microphone entitlement the hardened runtime requires. To distribute, set macos.signingIdentity (Developer ID) and macos.notarize in mygo.json; see mygo's distribution guide.

Try it without speaking

Replay a recording through Scribe and Translator, headless, with a JSON report:

read -r -s ZOOM_API_KEY && export ZOOM_API_KEY
go run . -replay e2e/testdata/captions-en.wav -lang-a en-US -lang-b zh-CN -translate -report /tmp/capy.json

Development

Capy is one Go binary built with mygo's native UI: no cgo, no webview, no server.

make test      # unit + integration (fake Zoom API, real SQLite)
make lint      # gofmt, go vet
make e2e       # replay fixtures against the fake API; reports must match e2e/testdata/*.golden.json
make e2e-live  # same fixtures against the real API (needs ZOOM_API_KEY)

Contributors: read AGENTS.md for the rules and package map.

License

MIT