Granola MCP Server
Integrates local Granola.ai meeting intelligence with Claude Desktop to enable searching and analyzing meeting transcripts, notes, and summaries. Users can perform natural language queries to retrieve meeting details, analyze participant patterns, and access full speaker-identified conversations.
README
Granola MCP Server (Experimental)
An experimental Model Context Protocol (MCP) server for integrating Granola.ai meeting intelligence with Claude Desktop. This uses Granola's local cache. I don't have any idea how Granola updates or maintains that cache. My guess is that it's doing a rolling context window, but storing long term notes up in AWS. So, YMMV. Use at your own risk. I will likely add a cache shipper at some point since we don't have access to Gronala's data in the cloud.
Features
- Meeting Search: Search meetings by title, content, participants, and transcript content
- Meeting Details: Get comprehensive meeting metadata with local timezone display
- Full Transcript Access: Retrieve complete meeting conversations with speaker identification
- Rich Document Content: Access actual meeting notes, summaries, and structured content
- Pattern Analysis: Analyze patterns across meetings (participants, frequency, topics)
- Timezone Intelligence: All timestamps automatically display in your local timezone
- Real-time Integration: Seamlessly connects to your actual Granola meeting data
Quick Start
Prerequisites
- Python 3.12+
- uv package manager
- macOS with Granola.ai installed
- Claude Desktop application
- Granola cache file at
~/Library/Application Support/Granola/cache-v3.json
Installation
-
Clone the repository to your home directory:
cd ~ git clone <repository-url> cd granola-ai-mcp-serverImportant: Clone to your home directory (
~) rather than~/Documentsto avoid macOS permission issues with Claude Desktop. -
Install dependencies with uv:
uv sync -
Test the installation:
uv run python test_server.py -
Configure Claude Desktop by adding to your
claude_desktop_config.json:{ "mcpServers": { "granola": { "command": "/Users/YOUR_USERNAME/granola-ai-mcp-server/.venv/bin/granola-mcp-server", "args": [], "env": {} } } }Important:
- Replace
YOUR_USERNAMEwith your actual macOS username - Use the direct path to the virtual environment's script (not
uv run) to avoid working directory issues - The path should point to your home directory installation
- Replace
-
Restart Claude Desktop to load the MCP server:
# Quit Claude completely osascript -e 'quit app "Claude"' # Reopen Claude open -a "Claude"
Available Tools
search_meetings
Search meetings by query string.
Parameters:
- query (string): Search query for meetings
- limit (integer, optional): Maximum number of results (default: 10)
get_meeting_details
Get detailed information about a specific meeting.
Parameters:
- meeting_id (string): Meeting ID to retrieve details for
get_meeting_transcript
Get transcript for a specific meeting.
Parameters:
- meeting_id (string): Meeting ID to get transcript for
get_meeting_documents
Get documents associated with a meeting.
Parameters:
- meeting_id (string): Meeting ID to get documents for
analyze_meeting_patterns
Analyze patterns across multiple meetings.
Parameters:
- pattern_type (string): Type of pattern to analyze ('topics', 'participants', 'frequency')
- date_range (object, optional): Date range for analysis with start_date and end_date
Usage Examples
Once configured with Claude Desktop, you can use natural language to interact with your Granola meetings:
Basic Queries
- "Search for meetings about quarterly planning"
- "Show me yesterday's meetings"
- "Find meetings with David from this week"
Transcript Access
- "Get the transcript from yesterday's ProofChat meeting"
- "What was discussed in the Float rollback planning meeting?"
- "Show me the full conversation from the David Tibbi meeting"
Content Analysis
- "Analyze participant patterns from last month"
- "What documents are associated with the product review meeting?"
- "Search for mentions of 'schema labeling' in meeting transcripts"
Recent Meeting Intelligence
The server automatically detects and provides access to:
- Full transcripts from recent meetings (25,000+ characters)
- Meeting content including notes and summaries
- Participant information and speaker identification
- Local timezone display for all meeting times
Development
Running Tests
uv run python test_server.py
Running the Server Directly
uv run granola-mcp-server
Adding Dependencies
uv add package-name
Configuration
Claude Desktop Config Locations
- macOS:
~/Library/Application Support/Claude/claude_desktop_config.json - Windows:
%APPDATA%\Claude\claude_desktop_config.json - Linux:
~/.config/Claude/claude_desktop_config.json
Granola Cache Location
The server reads from Granola's cache file at:
~/Library/Application Support/Granola/cache-v3.json
Security & Privacy
- ✅ 100% Local Processing - All data stays on your machine
- ✅ No External API Calls - No data sent to external services
- ✅ Granola Permissions Respected - Uses existing Granola.ai access controls
- ✅ Read-Only Access - Server only reads from Granola's cache
Performance & Capabilities
- Fast Loading: Sub-2 second cache loading for hundreds of meetings
- Rich Content: Extracts 25,000+ character transcripts and meeting notes
- Efficient Search: Multi-field search across titles, content, participants, and transcripts
- Memory Optimized: Lazy loading with intelligent content parsing
- Timezone Smart: Automatic local timezone detection and display
- Production Ready: Successfully processes real Granola data (11.7MB cache files)
- Scalable: Handles large datasets with 500+ transcript segments per meeting
Current Status
🚀 PRODUCTION READY - Successfully tested with real Granola.ai data including:
- ✅ 39+ meetings parsed and searchable
- ✅ 28 full transcripts with complete conversations
- ✅ Rich meeting content from notes, summaries, and structured data
- ✅ Timezone intelligence showing times like "17:04" instead of "21:04 UTC"
- ✅ Speaker identification and conversation flow
- ✅ Yesterday's meetings fully accessible with detailed content
Troubleshooting
Common Issues
"Cache file not found"
- Ensure Granola.ai is installed and has processed some meetings
- Check that the cache file exists:
ls -la "~/Library/Application Support/Granola/cache-v3.json"
"uv command not found"
- Install uv:
curl -LsSf https://astral.sh/uv/install.sh | sh - Or use pip fallback in Claude config:
"command": "python"
"Permission denied" or "Operation not permitted"
-
Most common issue on macOS: This happens when the server is installed in
~/Documentsor other protected folders -
Solution 1 (Recommended): Move the installation to your home directory:
mv ~/Documents/granola-ai-mcp-server ~/granola-ai-mcp-server cd ~/granola-ai-mcp-server uv sync # Rebuild venv with correct pathsThen update the path in
claude_desktop_config.json -
Solution 2: Grant Claude Desktop Full Disk Access:
- Open System Settings → Privacy & Security → Full Disk Access
- Click the lock icon and authenticate
- Click "+" and add
/Applications/Claude.app - Toggle Claude to "On"
- Restart Claude Desktop
"Current directory does not exist"
- This error occurs when using
uv runwith the--directoryflag - Use the direct path to the venv script instead (see Installation step 4)
Server not appearing in Claude Desktop
- Verify the absolute path in your Claude config
- Check Claude Desktop logs:
~/Library/Logs/Claude/mcp-server-granola.log - Look for Python errors in the logs
- Ensure the path doesn't contain spaces or special characters
- Restart Claude Desktop after config changes
"Failed to spawn process" or "No such file or directory"
- The Python shebang in the venv script points to the wrong location
- Run
uv syncin the project directory to rebuild the venv - Verify the script exists:
ls -la ~/granola-ai-mcp-server/.venv/bin/granola-mcp-server
Meeting notes appear empty in Claude
- Granola sometimes stores rich notes inside
documentPanelsrather thannotes_plain - This server now reads those panels by default; set
GRANOLA_PARSE_PANELS=0in the environment to disable - Run
python test_real_cache.pyto verify that panel-backed notes produce content
推荐服务器
Baidu Map
百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。
Playwright MCP Server
一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。
Magic Component Platform (MCP)
一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。
Audiense Insights MCP Server
通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。
VeyraX
一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。
graphlit-mcp-server
模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。
Kagi MCP Server
一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。
e2b-mcp-server
使用 MCP 通过 e2b 运行代码。
Neon MCP Server
用于与 Neon 管理 API 和数据库交互的 MCP 服务器
Exa MCP Server
模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。