zoom-summaries
Enables users to fetch Zoom meeting summaries and write them into matching Obsidian 1:1 notes, with browser-based authentication and automatic session cookie management.
README
Zoom Meeting Summaries MCP Server
An MCP (Model Context Protocol) server that provides access to Zoom meeting summaries and writes them into an Obsidian vault. Uses Puppeteer-based browser authentication — on first use, a browser window opens for you to sign in to Zoom. Session cookies are saved to ~/.zoom-mcp/cookies.json and reused automatically until they expire.
Configurable for any Zoom organization (e.g., acme.zoom.us, yourcompany.zoom.us).
Prerequisites
- Node.js 18+
- A Zoom account with AI Companion meeting summaries enabled
- An Obsidian vault with 1:1 note files named
First Last.md
Setup
npm install
npm run build
Configuration
| Variable | Description | Default |
|---|---|---|
ZOOM_SUBDOMAIN |
Your org's Zoom vanity subdomain (e.g. acme for acme.zoom.us) |
(empty — uses zoom.us) |
OBSIDIAN_VAULT_PATH |
Absolute path to your Obsidian vault root | (required) |
VAULT_SUBFOLDER |
Subfolder within the vault containing 1:1 note folders (e.g. MyOrg) |
(empty — search directly under vault root) |
ONE_ON_ONE_FOLDERS |
Comma-separated list of folder names holding 1:1 notes | ! One on Ones, ! One on Ones (Other) |
SHARED_MEETINGS_FOLDER |
(Optional, v1.1.0+) Folder for meetings shared WITH you. If not set, shared meetings are not processed. | (empty) |
Usage with Claude Desktop
Add to your Claude Desktop config (~/Library/Application Support/Claude/claude_desktop_config.json):
{
"mcpServers": {
"zoom-summaries": {
"command": "node",
"args": ["/absolute/path/to/zoomFetchMCP/build/index.js"],
"env": {
"ZOOM_SUBDOMAIN": "acme",
"OBSIDIAN_VAULT_PATH": "/path/to/your/vault",
"VAULT_SUBFOLDER": "MyOrg",
"ONE_ON_ONE_FOLDERS": "! One on Ones,! One on Ones (Other)"
}
}
}
}
Usage with VS Code Copilot (Agent Mode)
Ensure chat.agent.enabled is turned on in VS Code settings. Then add .vscode/mcp.json to your workspace:
{
"servers": {
"zoom-summaries": {
"type": "stdio",
"command": "node",
"args": ["/absolute/path/to/zoomFetchMCP/build/index.js"],
"env": {
"ZOOM_SUBDOMAIN": "acme",
"OBSIDIAN_VAULT_PATH": "/path/to/your/vault",
"VAULT_SUBFOLDER": "MyOrg",
"ONE_ON_ONE_FOLDERS": "! One on Ones,! One on Ones (Other)"
}
}
}
}
Available Tools
zoom_login
Opens a browser window for you to sign into Zoom. Call this first if you get authentication errors. Session cookies are saved to ~/.zoom-mcp/cookies.json for automatic reuse.
list_meeting_summaries
List meeting summaries with optional source and date filtering.
| Parameter | Type | Description |
|---|---|---|
source |
string (optional) | owned (default) or shared |
from |
string (optional) | Start date (e.g. 2025-01-01) |
to |
string (optional) | End date (e.g. 2025-02-01) |
get_meeting_summary
Get the full AI-generated summary for a specific meeting.
| Parameter | Type | Description |
|---|---|---|
meeting_id |
string | The Zoom meeting ID |
write_summary_to_obsidian
Fetch a Zoom meeting summary and write it into the matching Obsidian 1:1 note. Matches the note file by the first name in the meeting topic (e.g. Amit:Howard → Amit Kumar.md). After a successful write, deletes the summary from Zoom.
| Parameter | Type | Description |
|---|---|---|
meeting_id |
string | The numeric Zoom meeting ID (e.g. 96980348286) |
write_all_summaries_to_obsidian
Fetch every Zoom meeting summary and write each one to the matching Obsidian 1:1 note. Skips multi-person meetings (not 1:1s), skips meetings with no matching file, and skips duplicates already present in the note. After each successful write (including duplicates already in the vault), deletes the summary from Zoom.
| Parameter | Type | Description |
|---|---|---|
from |
string (optional) | Start date filter (e.g. 2026-01-01) |
to |
string (optional) | End date filter (e.g. 2026-02-23) |
Obsidian Integration
File matching
The tool looks for a vault note to write each summary into using this priority:
- Exact attendee name match — a file named
First Last.mdmatching an attendee - Topic first-name prefix — file whose name starts with the first name from the topic (e.g.
AmitfromAmit:Howard), preferring solo files over shared ones (Amit Kumar.mdoverAmit Kumar + Bob.md) - Word-boundary search — broader name search as fallback
Multi-person meetings (3+ participants, or Name1:Name2:Name3 topic format) are skipped — only 1:1s are written.
Note format
Summaries are inserted in the vault's tab-indented style, newest-first:
- If the date already has manual notes: Zoom content is appended below the existing notes for that day (with a blank line separator), without repeating the date header. Starts with
\tZoom AI Summary. - If the date is new: A standalone
YYYY-MM-DD - Zoom AI Summarysection is inserted at the correct chronological position.
Caches
All caches are in-memory only and scoped to a single run. They exist solely to avoid fetching the same meeting twice within one execution (e.g. a meeting whose attendees were resolved in the attendee phase need not be re-fetched in the summary phase). No cache data is ever written to disk. ~/.zoom-mcp/ only holds cookies.json (session) and optionally config.json (vault path).
Dry-Run / Live Script
zoom-meetings-to-obsidian.mjs is a standalone Node.js script for previewing and applying changes.
# Dry-run: copies vault files to /tmp/obsidian-preview and applies mutations there
node zoom-meetings-to-obsidian.mjs
# Write to real vault (no deletion)
node zoom-meetings-to-obsidian.mjs --update
# Write to real vault and delete summaries from Zoom
node zoom-meetings-to-obsidian.mjs --update --delete
# Add --debug to any command for verbose diagnostics (page elements, network requests, etc.)
node zoom-meetings-to-obsidian.mjs --update --delete --debug
--delete without --update is an error.
Output includes:
- A table of every meeting: topic, target file, and insert/duplicate/skip status
- A
Would delete from Zoomlist (dry-run) or live deletion results (--do-it)
Inspect dry-run results in /tmp/obsidian-preview before running with --do-it.
How It Works
- On first use (or when cookies expire), calling
zoom_loginopens a real browser window - You sign in to Zoom normally (SSO, password, etc. — any method works)
- Session cookies are saved to
~/.zoom-mcp/cookies.json - Subsequent calls use a headless browser with saved cookies to scrape the summaries page
- Meeting summaries are written into Obsidian notes using chronological insertion
- After writing, the summary is deleted from Zoom's AI Companion summary list
- If cookies expire, call
zoom_loginagain
Development
npm run build # Compile TypeScript → build/
npm run dev # Watch mode (tsc --watch)
npm start # Run the MCP server (stdio transport)
Recommended Servers
playwright-mcp
A Model Context Protocol server that enables LLMs to interact with web pages through structured accessibility snapshots without requiring vision models or screenshots.
Magic Component Platform (MCP)
An AI-powered tool that generates modern UI components from natural language descriptions, integrating with popular IDEs to streamline UI development workflow.
Audiense Insights MCP Server
Enables interaction with Audiense Insights accounts via the Model Context Protocol, facilitating the extraction and analysis of marketing insights and audience data including demographics, behavior, and influencer engagement.
VeyraX MCP
Single MCP tool to connect all your favorite tools: Gmail, Calendar and 40 more.
graphlit-mcp-server
The Model Context Protocol (MCP) Server enables integration between MCP clients and the Graphlit service. Ingest anything from Slack to Gmail to podcast feeds, in addition to web crawling, into a Graphlit project - and then retrieve relevant contents from the MCP client.
Kagi MCP Server
An MCP server that integrates Kagi search capabilities with Claude AI, enabling Claude to perform real-time web searches when answering questions that require up-to-date information.
Neon Database
MCP server for interacting with Neon Management API and databases
E2B
Using MCP to run code via e2b.
Exa Search
A Model Context Protocol (MCP) server lets AI assistants like Claude use the Exa AI Search API for web searches. This setup allows AI models to get real-time web information in a safe and controlled way.
Qdrant Server
This repository is an example of how to create a MCP server for Qdrant, a vector search engine.