自定义MCP服务器将 Granola.ai 会议数据连接到Claude Desktop，实现AI驱动的会议智能
概览
一个定制的模型上下文协议（MCP）服务器，将 Granola.ai 会议数据直接连接到Claude Desktop，实现对整个会议历史的自然语言查询。 虽然 Granola-MCP 在网络上有详细的功能列表和安装指南文档，我们没有等待正式发布，而是逆向工程了 Granola 的本地缓存结构，从零开始构建了一个全面的 MCP 服务器。 问Claude一些问题，比如：“给我看上个月所有讨论预算的客户会议”- “哪些团队成员需要安然项目会议的跟进？
概述（AI 翻译）
将您的会议数据转化为AI驱动的商业智能
一个定制的模型上下文协议（MCP）服务器，将 Granola.ai 会议数据直接连接到Claude Desktop，实现对整个会议历史的自然语言查询。
🚀 为什么会有这样的存在
虽然Granola-MCP在网络上有详细的功能列表和安装指南，但实际的仓库和PyPI包并不存在。我们没有等待正式发布，而是逆向工程Granola的本地缓存结构，从零开始构建了一个全面的MCP服务器。
结果如何？ 问克劳德诸如此类问题：
- “给我看上个月所有讨论预算的客户会议”
- “哪些团队成员需要安然项目会议的跟进？”
- “分析我们与外部客户的会议模式与内部团队会议的模式”
✨ 特色
🔍 自然语言与智能相遇
- 使用通俗英语查询273+会议
- 跨标题、内容和参与者列表进行搜索
- 即时访问会议元数据和日历集成
📊 高级分析
- 会面统计与模式识别
本页由 AI 翻译，技术细节请以原文为准。
README
以下为原文 README 节选，完整内容请查看原文。
Granola-Claude MCP Server
Transform your meeting data into AI-powered business intelligence
A custom Model Context Protocol (MCP) server that connects Granola.ai meeting data directly to Claude Desktop, enabling natural language queries across your entire meeting history.
🚀 Why This Exists
While Granola-MCP was documented across the web with detailed feature lists and installation guides, the actual repository and PyPI package didn’t exist. Rather than wait for an official release, we reverse-engineered Granola’s local cache structure and built a comprehensive MCP server from scratch.
The result? Ask Claude questions like:
- “Show me all client meetings from last month where we discussed budget”
- “Which team members need follow-up from the Enron project meetings?”
- “Analyze our meeting patterns with external clients vs internal team meetings”
✨ Features
🔍 Natural Language Meeting Intelligence
- Query 273+ meetings using plain English
- Search across titles, content, and participant lists
- Instant access to meeting metadata and calendar integration
📊 Advanced Analytics
- Meeting statistics and pattern recognition
- Attendee analysis and engagement tracking
- Project timeline reconstruction from meeting history
🔗 Complete Integration
- Google Calendar event details and attendee responses
- Conference links (Google Meet, Zoom) and participation status
- Rich metadata including organizers, response status, and sharing settings
🛡️ Privacy-First Architecture
- All data processing happens locally - no cloud uploads
- Direct cache file access - no API dependencies
- Respects existing Granola permissions and access controls
🎯 Use Cases
Project Management
"Show meetings about the XXI Media project and identify next steps"
→ Surfaces July 16 Zoom meeting with Pyxl, attendee status, action items
Client Relations
"Which clients had 'needsAction' responses this week?"
→ Identifies follow-up requirements across all client touchpoints
Team Analytics
"Who attends the most external client meetings?"
→ Analyzes workload distribution and development opportunities
Meeting ROI
"Compare meeting outcomes between Google Meet vs Zoom sessions"
→ Platform effectiveness analysis for future planning
🛠️ Installation
Prerequisites
- macOS with Granola.ai installed and configured
- Python 3.12+
- Claude Desktop application
- Active Granola meeting history
Quick Start
- Clone the repository
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": {
"granola-mcp": {
"command": "python3",
"args": ["/path/to/granola-claude-mcp/granola_mcp_server.py"],
"env": {
"GRANOLA_CACHE_PATH": "/Users/[username]/Library/Application Support/Granola/cache-v3.json"
}
}
}
}