MCP server for accessing Spark Desktop meeting transcripts and emails through the Model Context Protocol.
概览
MCP server for accessing Spark Desktop meeting transcripts and emails through the Model Context Protocol. This server is installed and registered for Claude Code and Claude Desktop by setup-mcp.sh in the dotfiles repo, along with all the other MCP servers: Then restart Claude Code and Claude Desktop. (More detail, including upgrading an older Mac: the laptop-setup README.) The script clones or pulls every server repo into ~/src, builds each environment from its uv.lock, writes the config, and checks that each server starts. It's safe to re-run anytime. After that, a plain git pull in this repo is enough: the server launches with uv run --frozen, which installs any changed dependencies on the next start. Credentials live in ~/.mcp-auth (synced from Box with unison mcp_auth).
README
以下为原文 README 节选,完整内容请查看原文。
Spark MCP Server
MCP server for accessing Spark Desktop meeting transcripts and emails through the Model Context Protocol.
Quick setup / update (any Mac)
This server is installed and registered for Claude Code and Claude Desktop by
setup-mcp.sh in the dotfiles repo,
along with all the other MCP servers:
cd ~/dotfiles && git pull && ./setup-mcp.sh
Then restart Claude Code and Claude Desktop. (More detail, including upgrading an older Mac: the
laptop-setup README.)
The script clones or pulls every server repo into
~/src, builds each environment from its uv.lock, writes the config, and checks that each server
starts. It’s safe to re-run anytime.
After that, a plain git pull in this repo is enough: the server launches with uv run --frozen,
which installs any changed dependencies on the next start. Credentials live in ~/.mcp-auth
(synced from Box with unison mcp_auth).
Features
- 📝 Access all meeting transcripts (calendar-based and ad-hoc)
- 🔍 Full-text search across transcript content
- 📊 Statistics and analytics about your transcripts
- 🔒 Read-only access - safe and non-destructive
- ⚡ Fast local SQLite queries - no network required
- 🎯 Captures ad-hoc meetings (primary use case)
Requirements
- macOS (Spark Desktop must be installed)
- Python 3.10+
- Spark Desktop for macOS (App Store version)
Installation
# Install in development mode
pip install -e .
Usage
With Claude Desktop
Add to your Claude Desktop MCP settings (~/Library/Application Support/Claude/claude_desktop_config.json):
{
"mcpServers": {
"spark": {
"command": "python",
"args": ["-m", "spark_mcp.server"],
"cwd": "/Users/feamster/src/spark-mcp"
}
}
}
Or if installed via pip:
{
"mcpServers": {
"spark": {
"command": "spark-mcp"
}
}
}
Restart Claude Desktop, and the tools will be available.
Standalone Testing
# Run the server (communicates via stdio)
python -m spark_mcp.server
Available Tools
1. list_meeting_transcripts
List meeting transcripts with metadata.
Parameters:
limit(optional, default: 20): Max resultsafter(optional): Return transcripts with meetingStartDate after this ISO datetime (e.g., ‘2026-01-30T13:00:00’)before(optional): Return transcripts with meetingStartDate before this ISO datetime (e.g., ‘2026-01-30T16:00:00’)
Example:
{
"after": "2026-02-09T09:00:00",
"before": "2026-02-09T12:00:00",
"limit": 10
}
Returns:
{
"transcripts": [
{
"messagePk": 63336,
"subject": "Prior Art Review for Patent Claims 416 and 571",
"sender": "[email protected]",
"receivedDate": "2025-11-11 15:59:12",
"meetingStartDate": "2025-11-11T15:00:00.000Z",
安装
This server does not publish a one-line install command.
Open the repository installation guide 新手?先看 mcpskill.org 教程 →配置
{
"mcpServers": {
"spark": {
"command": "python",
"args": ["-m", "spark_mcp.server"],
"cwd": "/Users/feamster/src/spark-mcp"
}
}
}