Content
# Local-Data-MCP-Connector
An open-source, highly secure, and offline-first **Model Context Protocol (MCP) Server** designed to give AI agents (like Claude Desktop, Antigravity, or Cursor) read-only access to query your local calendar, email, and CSV logs.
This allows your local AI assistant to help you plan meetings, summarize client emails, and analyze fitness/financial logs without uploading any of your private files to the cloud.
---
## 🔒 Security & Privacy
1. **100% Offline Parsing:** All processing is done locally on your CPU using Python standard libraries.
2. **Strictly Read-Only:** The server only exposes query/read tools. It cannot modify, delete, or create local files.
3. **Automatic Git Protection:** The project's `.gitignore` is pre-configured to block calendar (`.ics`), email (`.eml`/`.mbox`), and CSV data files from ever being accidentally committed to your public GitHub profile.
4. **Zero Dependencies:** Runs on standard vanilla Python libraries (no `pip install` required for runtime, making it highly secure and lightweight).
---
## 🚀 Getting Started
### 1. Prerequisites
- **Python 3.10+**
### 2. Clone the Repository
```bash
git clone https://github.com/yourusername/local-data-mcp-connector.git
cd local-data-mcp-connector
```
### 3. Run Automated Unit Tests
Verify the parser works correctly on synthetic test datasets:
```bash
python test_connector.py
```
---
## 🛠️ Registering the MCP Server
Configure your AI desktop client to execute this server via stdio.
### For Claude Desktop
Add this to `%APPDATA%\Claude\claude_desktop_config.json` (Windows):
```json
{
"mcpServers": {
"local-data-mcp-connector": {
"command": "python",
"args": [
"C:\\path\\to\\local-data-mcp-connector\\mcp_connector.py"
]
}
}
}
```
*(Make sure to replace the path with the absolute path where you cloned the repo)*
### For Google Antigravity
Add this to `~/.gemini/config/mcp_config.json`:
```json
{
"mcpServers": {
"local-data-mcp-connector": {
"command": "python",
"args": [
"C:\\path\\to\\local-data-mcp-connector\\mcp_connector.py"
]
}
}
}
```
---
## 🛠️ Exposed MCP Tools
The server exposes the following tools:
1. `query_calendar(file_path, query)`: Reads local `.ics` calendar files, searching event summaries and descriptions.
2. `query_email(file_path, query)`: Loads `.eml` emails, returning subjects, senders, dates, and plain-text summaries.
3. `query_mbox_archive(file_path, query)`: Scans `.mbox` local folders (used by Thunderbird, Google Takeout, etc.) to list matching emails.
4. `query_csv_log(file_path, query_column, query_value, row_limit)`: Queries rows inside local `.csv` spreadsheets (e.g. fitness data, bank exports).
