google_meet

Joins Google Meet calls with a Playwright bot to transcribe captions and speak via OpenAI Realtime.

Updated May 18, 2026
One-click install
npx skills add https://github.com/ossoolli/Nexum-Core --skill google-meet-ossoolli
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: google_meet
Source: https://github.com/ossoolli/Nexum-Core/tree/main/plugins/google_meet
Command: npx skills add https://github.com/ossoolli/Nexum-Core --skill google-meet-ossoolli

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill requires playwright, websockets.

What problem does it solve? Attending and documenting Google Meet calls manually is time-consuming, and agents normally cannot join video calls, capture what was said, or speak in them. This Skill lets an agent join a Meet URL, scrape live captions into a transcript, optionally speak in the call, and handle follow-up work like summaries and action items. ## Core Features & Use Cases - Live Transcription: A headless Chromium bot joins the meeting, enables Google Meet captions, and scrapes them into a timestamped transcript file via a DOM MutationObserver. - Realtime Voice: In realtime mode, the agent speaks into the call through OpenAI Realtime, with audio routed into Chrome's fake microphone via a PulseAudio null-sink (Linux) or BlackHole (macOS), including best-effort barge-in when humans interrupt. - Remote Node Hosts: The bot can run on a separate machine (e.g., a Mac with signed-in Chrome) controlled from a headless Linux gateway over a token-authenticated WebSocket connection. - Use Case: Ask the agent to sit in on your standup, take notes, and afterwards send a recap email with action items extracted from the full transcript. ## Quick Start Ask the agent to join your meeting by saying "join my Meet at https://meet.google.com/abc-defg-hij and take notes", after running hermes meet install once to set up Playwright and Chromium.

Frequently Asked Questions about google_meet

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I make an AI agent join a Google Meet call?▼

Call the meet_join tool with the meeting URL, or run hermes meet join <url> from the CLI. A headless Chromium bot joins the call, enables live captions, and writes a transcript file you can poll with meet_status and meet_transcript.

How can a bot speak in a Google Meet call?▼

Join with mode='realtime', then call meet_say with the text to speak. The bot sends text to OpenAI Realtime and streams the generated PCM audio into Chrome's fake microphone through a PulseAudio null-sink on Linux or BlackHole on macOS.

Does the Google Meet bot work on Windows?▼

No, Windows is not supported. The plugin only registers on Linux and macOS because the realtime audio bridging path is untested on Windows and guest-join Chromium is flakier there.

Why is the Meet bot stuck in the lobby?▼

Guest-mode joins wait in the lobby until a host admits the bot. If admission does not happen within 5 minutes (configurable via HERMES_MEET_LOBBY_TIMEOUT), the bot leaves with leaveReason 'lobby_timeout'. Running hermes meet auth to save a signed-in Google session avoids the lobby wait.

Can the bot run on a different machine than the agent?▼

Yes, using remote node hosts. Run hermes meet node run on the machine with signed-in Chrome, approve it on the gateway with hermes meet node approve, then pass node='<name>' to any meet tool to operate the remote bot over a token-authenticated WebSocket.

What are the limitations of Google Meet caption transcription?▼

Transcripts are only as accurate as Google Meet's built-in live captions, which are English-biased and lossy on overlapping speakers. The bot scrapes caption DOM elements rather than parsing WebRTC audio, so no STT API keys or billing are required.