Wispy Meet

See it work

Recorded from the running app: two questions answered from a private notes vault, then the theme picker.

What it does

Works with

Wispy never joins your call. It reads the Windows microphone state and the system audio you already hear, so any app that uses your mic is covered, including tabs in a browser.

ZoomGoogle MeetMMicrosoft TeamsSSlack huddlesDiscordWebexJitsiGoTo MeetingWhatsApp callsTelegram callsAny browser tab

Runs on any model

Pick a provider in Settings. With Ollama or LM Studio nothing leaves your machine at all.

Claude CLIyour subscription, no key

Anthropic APIsk-ant key

GGroqfastest tokens

Ollamafully offline

LM Studiofully offline

OpenRouterany open model

Set up in one prompt

Install Wispy, then paste this into Claude Code. It wires Wispy to your notes folder and adds a Wispy Meet segment to your Claude status line without touching the one you already have.

Set up Wispy Meet on this machine and wire it into my Claude Code setup.

1. Confirm the Claude CLI is installed by running: claude --version
   If it is missing, stop and tell me how to install it.

2. Find my Wispy config at %APPDATA%Wispyconfig.json
   If it does not exist, tell me to launch Wispy once so it is created, then stop.

3. Ask me which folder of Markdown notes Wispy should use as context.
   Set "contextPath" to that folder.
   Set "notesPath" to <that folder>meetings unless I say otherwise.

4. Set "providerId" to "claude-cli" and "backend" to "cli" so Wispy answers
   through my Claude subscription and needs no API key.

5. Add a Wispy Meet segment to my Claude Code status line:
   - Wispy writes live state to %APPDATA%Wispystate.json with fields
     meeting (bool), app (string), turns (number) and at (ISO timestamp).
   - If ~/.claude/settings.json has no statusLine, create a small script and
     point statusLine at it. It should print "Wispy Meet", plus the meeting app
     and question count when a meeting is live, and stay quiet when state.json
     is older than 10 minutes.
   - If a statusLine already exists, append the Wispy segment to that existing
     script. Do not replace or overwrite my current status line.

6. Verify by printing the resolved contextPath, notesPath and backend, then run
   the status line once so I can see the Wispy Meet segment.

Show me a diff before writing to settings.json or any existing script.

Setup by hand

1. 1Run the installer. It installs for your user only, so there is no admin prompt.
2. 2Wispy starts in the system tray and begins indexing your notes vault. The first index takes a while; leave it running.
3. 3Open Settings from the tray and point 'Context folder' at any folder of Markdown notes - an Obsidian vault works as-is.
4. 4Join any call. The overlay appears on its own. Press Ctrl+Shift+Space to answer the last thing you heard.

Shortcuts

Privacy

Speech is transcribed on your own machine and audio is never uploaded. When you ask a question, Wispy sends only that question, a rolling transcript excerpt and the matching notes to the model you choose, and with Ollama or LM Studio even that stays on your machine. Recording other people in a call can require their consent depending on where you and they are, so meeting audio capture is a setting you control, and you can turn it off and have Wispy hear only you.