SF Cleaning Service
# IRL - SF Cleaning Service
IRL - A Model Context Protocol (MCP) server for booking cleaning services in San Francisco. Can be deployed locally or to Vercel for remote access.
## Setup
### 1. Install Dependencies
```bash
npm install
```
### 2. Set Up Resend API
1. **Create Resend Account**: Go to https://resend.com and sign up
2. **Get API Key**:
- Navigate to your Resend Dashboard
- Go to "API Keys" section
- Create a new API key or copy your existing one
- It will look like: `re_BcASVtoX_Bj4QhZei4xSjyyLr21vhMbVd`
3. **Verify Domain (Required for custom email)**:
- To send from gwen@irl-concierge.com
- Go to "Domains" in Resend Dashboard
- Add domain: irl-concierge.com
- Follow DNS verification steps (add TXT, MX records)
- Once verified, you can send from gwen@irl-concierge.com
### 3. Configure Email Credentials
Edit `.env` file with your Resend credentials:
```
RESEND_API_KEY=re_xxxxxxxxxxxxxxxxxx
FROM_EMAIL=gwen@irl-concierge.com
PARTNER_EMAILS=partner1@example.com,partner2@example.com
```
**Note:** You must verify the domain `irl-concierge.com` in Resend before you can send from `gwen@irl-concierge.com`. Until verified, use `onboarding@resend.dev` for testing.
### 4. Configure Claude Desktop
Add to your Claude Desktop configuration file:
**macOS:** `~/Library/Application Support/Claude/claude_desktop_config.json`
**Windows:** `%APPDATA%\Claude\claude_desktop_config.json`
```json
{
"mcpServers": {
"sf-cleaning": {
"command": "node",
"args": ["/absolute/path/to/mcp-sf-cleaning/index.js"]
}
}
}
```
Replace `/absolute/path/to/mcp-sf-cleaning/index.js` with the actual path to your index.js file.
### 5. Restart Claude Desktop
After configuration, restart Claude Desktop to load the MCP server.
## Vercel Deployment (For Remote Access)
### 1. Install Vercel CLI
```bash
npm install -g vercel
```
### 2. Set Environment Variables
```bash
cd /Users/rana/Downloads/IRL/mcp-sf-cleaning
# Set up Vercel environment variables
vercel env add RESEND_API_KEY
# Enter: re_BcASVtoX_Bj4QhZei4xSjyyLr21vhMbVd
vercel env add FROM_EMAIL
# Enter: gwen@irl-concierge.com
vercel env add PARTNER_EMAILS
# Enter: ranadaytoday@outlook.com
```
### 3. Deploy to Vercel
```bash
vercel --prod
```
Follow the prompts:
- Set up and deploy? Y
- Which scope? (select your account)
- Link to existing project? N
- Project name? irl
- Directory? ./
- Want to modify settings? N
### 4. Get Your MCP Server URL
After deployment, you'll get a URL like:
```
https://mcp-sf-cleaning.vercel.app
```
### 5. Add to Claude Desktop as Custom Connector
1. Open Claude Desktop
2. Go to Settings → Connectors
3. Click "Add custom connector"
4. Enter:
- **Name**: SF Cleaning Service
- **Remote MCP server URL**: `https://mcp-sf-cleaning.vercel.app/api/mcp`
5. Save and restart Claude
### Testing the Remote Server
You can test the API directly:
```bash
# List tools
curl -X POST https://mcp-sf-cleaning.vercel.app/api/mcp \
-H "Content-Type: application/json" \
-d '{"method":"tools/list"}'
# Request cleaning (SF address)
curl -X POST https://mcp-sf-cleaning.vercel.app/api/mcp \
-H "Content-Type: application/json" \
-d '{
"method": "tools/call",
"params": {
"name": "request_cleaning",
"arguments": {
"name": "John Doe",
"phone": "415-555-1234",
"address": "123 Market St, SF 94105"
}
}
}'
```
## Test Conversation Examples
### Example 1: San Francisco Address (Accepted)
```
You: Can you book a cleaning service for me?
Claude: I'll help you book a cleaning service. I'll need your name, phone number, and address.
You: Name is John Doe, phone is 415-555-0123, address is 123 Market St, San Francisco, CA 94105
Claude: [Uses request_cleaning tool]
Result: ✅ Sent! John Doe, they'll call 415-555-0123 within 1 hour.
```
### Example 2: Non-SF Address (Rejected)
```
You: Book cleaning for Sarah Smith, 510-555-9876, 456 Broadway, Oakland, CA 94607
Claude: [Uses request_cleaning tool]
Result: Sorry, we only serve San Francisco currently. We're expanding - stay tuned!
```
### Example 3: SF Zip Code Detection
```
You: I need cleaning at 789 Pine Street, 94108. Name: Alice Wong, Phone: 415-555-3456
Claude: [Uses request_cleaning tool]
Result: ✅ Sent! Alice Wong, they'll call 415-555-3456 within 1 hour.
```
## Usage
The server provides one tool:
- `request_cleaning`: Books cleaning service for SF addresses only
### Parameters:
- `name`: Customer name
- `phone`: Contact phone number
- `address`: Service address (must be in San Francisco)
### Responses:
- **SF Address**: "✅ Sent! [name], they'll call [phone] within 1 hour."
- **Non-SF Address**: "Sorry, we only serve San Francisco currently. We're expanding - stay tuned!"
## How It Works
1. Checks if address contains "sf", "san francisco", or SF zip codes (940xx, 941xx)
2. If in SF: Sends email via Resend API to partners and confirms booking
3. If not in SF: Returns polite rejection message
## Why Resend?
- **Simple API**: Clean, modern email API designed for developers
- **No SMTP hassles**: No need for app passwords or complex SMTP settings
- **Better deliverability**: Built-in SPF, DKIM, and DMARC support
- **Free tier**: 3,000 emails/month free, perfect for small projectsTest auto-deployment
TDQS
Scored across 1 tool
With only one tool, there is no possibility of ambiguity or overlap between tools. The single tool 'request_cleaning' has a clear, distinct purpose that cannot be confused with any other tool in this set.
The naming pattern is trivially consistent since there is only one tool. The tool name 'request_cleaning' follows a clear verb_noun convention, and with no other tools to compare, there is no inconsistency.
A single tool is generally too few for a server's purpose, as it limits functionality and suggests an incomplete or thin surface. For a 'SF Cleaning Service' domain, one might expect additional tools for operations like scheduling, pricing, or status checks, making this count inappropriate for the apparent scope.
The tool set is severely incomplete for a cleaning service domain. While 'request_cleaning' covers initiation, there are obvious gaps such as checking service status, modifying or canceling requests, viewing pricing, or managing schedules. This will likely cause agent failures in handling full workflows.