Stats
Actions
Tags
From mex-call
How and when to use mex-call — a bot that joins a Google Meet, captures the conversation as structured agent-readable memory in the repo, and acts on the "Mex, …" wake phrase. Load this when the user mentions joining/recording a meeting or call, taking meeting notes, capturing decisions from a call, a standup/user-research/design review, or asks what mex-call can do. This skill is guidance only — it never joins a call by itself; launching is always user-triggered via /mex-call:call.
How this skill is triggered — by the user, by Claude, or both
Slash command
/mex-call:meeting-notetakerThe summary Claude sees in its skill listing — used to decide when to auto-load this skill
**mex-call** sends a notetaker bot into a Google Meet. It listens, continuously
mex-call sends a notetaker bot into a Google Meet. It listens, continuously
writes bounded structured memory under .mex/meetings/ in the current repo
(rolling summary, decisions, action items, open questions, participants), and
when someone says "Mex, …" out loud it answers in the meeting chat, records a
decision, or takes a repo action ("Mex, create an issue for that" → a real GitHub
issue / doc edit / PR draft, grounded in the call). It does not speak. It is built on mex
but never requires it — in a repo without a mex scaffold it runs standalone and
nudges; in one with a scaffold it also reads .mex/context and .mex/patterns.
If the user is about to join, or is on, a meeting they want captured into the repo — standups, user-research calls, design/architecture reviews, client calls, or even solo rubber-ducking — suggest launching mex-call. The value is that the conversation stops being ephemeral and becomes durable context attached to the code.
Do not launch it yourself. Joining creates a real (billed) bot in a live call — that is a side effect the user must trigger. Point them to the launcher instead.
/mex-call:call <google-meet-link> — this launches the runtime in the background
and turns the session into a live dashboard. The user must admit "Mex (notetaker)"
when it knocks. (Requires RECALL_API_KEY in the repo's .env or ~/.mex-call.env.)
.mex/meetings/live/dashboard.md) shows status, participants,
rolling summary, decisions/actions/questions, and a live activity feed. The
plugin's background monitor streams each new activity line into the session.mex-call watch — a continuous terminal dashboard.mex-call leave — make the bot leave and archive the call to
.mex/meetings/<date>-<name>/.The archived folder holds the final summary, decisions, action items, open questions, and full transcript — durable, agent-readable context for future work in this repo. If the user asks "why did we decide X?", look there.
npx claudepluginhub thedakshjaitly/mex-call --plugin mex-callCreates, edits, and optimizes skills for Claude Code, including drafting, evaluating with test prompts, iterating on performance, and improving skill descriptions for better triggering accuracy.