# Ops MCP Server

Real MCP (Model Context Protocol) WebSocket server for Gmail operations management. This connects to Agent Builder and provides tools for fetching ops reports and managing tasks.

## Quick Start

### 1. Setup
```bash
# Install dependencies
npm install

# Copy and configure environment
cp .env.example .env
# Edit .env and add your AI_SECRET_TOKEN
```

### 2. Run locally
```bash
npm start
# Server runs on ws://localhost:3334
```

### 3. Deploy to Railway

1. Push to GitHub:
```bash
git init
git add .
git commit -m "Initial MCP server"
git remote add origin https://github.com/johnohhh1/ops-mcp-server.git
git push -u origin main
```

2. Connect Railway to your GitHub repo
3. Add environment variable: `AI_SECRET_TOKEN`
4. Deploy! You'll get a URL like `wss://ops-mcp-server.up.railway.app`

### 4. Connect to Agent Builder

In OpenAI Agent Builder:
- Tools → Add → MCP
- URL: `wss://your-railway-app.up.railway.app`
- Tools auto-discover: `fetch_ops_report` and `mark_task_done`

## Architecture

```
Agent Builder
     ↓ (MCP WebSocket)
This MCP Server
     ↓ (HTTPS + Bearer)
Your Vercel API (my-ops-runner)
     ↓
Gmail (future)
```

## Available Tools

- `fetch_ops_report` - Gets HotSchedules 911 items and Brinker deadlines
- `mark_task_done` - Marks a task complete by ID

## Environment Variables

- `AI_SECRET_TOKEN` - Your secret token (must match Vercel)
- `PORT` - WebSocket port (default 3334)