Eventsorg MCP

🌐 EventsOrg MCP: Autonomous AI Event Manager

An open-source Model Context Protocol (MCP) server and background autonomous agent designed to run the logistics of live events.

EventsOrg MCP gives LLMs (like Claude and Gemini) the "hands" to read live event data, reassign volunteers, cancel missing vendors, and communicate with staff via physical SMS messages.

Note: This repository is a framework. It is designed so that any event organizer can deploy their own instance of this code, attach their own API keys, and manage their own isolated event.

---

✨ Features

• MCP Protocol Native: Exposes over 20+ event management tools (e.g., reassign_volunteer, get_vendor_status) to any standard AI client.
• Twilio SMS Orchestration: The AI can text volunteers physical SMS messages. Inbound SMS replies are automatically parsed and routed back into the database.
• Autonomous Agent Loop: Includes a standalone Gemini-powered agent (agent.ts) that runs 24/7, actively monitoring the event and resolving issues before a human even notices.
• Google Sheets Sync: Automatically pushes real-time database state to a Google Sheets dashboard for human coordinators to view.
• Zero-Config Local Webhooks: Uses Cloudflare Tunnels (untun) to automatically route live Twilio SMS back to your localhost during development.

---

🏗 Architecture & Use Cases

This codebase supports three entirely different ways to interact with the AI:

1. Local Desktop Plugin (stdio)

Perfect for solo event organizers or local development. By connecting the code to Claude Desktop or Cursor, the AI talks directly to the local SQLite database via standard terminal input/output (stdio).

• *Requires: Your laptop to stay open and running.*

2. Public API for AI (HTTP / SSE)

By starting the server with the --transport http flag, you turn the local server into a globally accessible web API. Other developers, LangChain apps, or AI IDEs can connect to your event remotely using Server-Sent Events (SSE).

• *Requires: Deployment to a cloud host like Railway, AWS, or Render.*

3. The Autonomous Virtual Coordinator

You don't even need a chat window. By running the included agent.ts script, a Gemini-powered worker wakes up every 5 minutes, checks the database for anomalies (missing vendors, safety incidents, weather alerts), fixes them, and texts the staff—completely autonomously.

4. Packaged Agent Skill (SKILL.md)

EventsOrg is packaged as a standard Agent Skill compliant with the open SKILL.md specification and listed on marketplaces like Agensi.io. It provides AI agents (Claude Code, Cursor, Copilot) with explicit SOPs, incident triage rules, and real-time SMS dispatch protocols out of the box.

• *Package folder:* skills/event-logistics-coordinator/

---

🚀 How to Deploy Your Own Event

Because this system uses a local SQLite database (grid.db), one deployment equals one event. If multiple organizers want to run different events, they must deploy their own separate instances of this code.

Prerequisites

You will need API keys for:

• Twilio (Account SID, Auth Token, Phone Number)
• Google Cloud (Service Account JSON with Sheets API enabled)
• Google Gemini (For the autonomous agent)

Environment Variables

Copy the .env.example file to .env and configure your API keys before starting the server. Alternatively, if deploying to a cloud platform (like Render, AWS, or Railway), add these directly to your dashboard:

# AI Models

GOOGLE_API_KEY=your_gemini_api_key

Twilio SMS Engine

TWILIO_ACCOUNT_SID=your_twilio_sid TWILIO_AUTH_TOKEN=your_twilio_token TWILIO_PHONE_NUMBER=+1234567890

External Integrations

OPENWEATHER_API_KEY=your_openweather_key GOOGLE_MAPS_API_KEY=your_maps_key

Escalation Emails

COORDINATOR_EMAIL=your_email@gmail.com SMTP_USER=your_email@gmail.com SMTP_PASS=your_app_password

Webhook Config

PUBLIC_URL=eventsorg-mcp-production.up.railway.app # Set this when deployed publicly WEBHOOK_PORT=3000

Google Sheets Sync

GOOGLE_SHEET_ID=your_google_sheet_id GOOGLE_SERVICE_ACCOUNT_JSON='{ "type": "service_account", ... }'

1. Using with Claude Desktop (stdio mode)

The easiest way to use this is as a local plugin for Claude. Since it's published to NPM, you don't even need to clone the repo.

1. Open your claude_desktop_config.json file.
2. Add this configuration (inserting your own API keys):

   {

"mcpServers": { "eventsorg-mcp": { "command": "npx", "args": ["-y", "eventsorg-mcp"], "env": { "TWILIO_ACCOUNT_SID": "your_twilio_sid", "TWILIO_AUTH_TOKEN": "your_twilio_token", "TWILIO_PHONE_NUMBER": "+1234567890", "GOOGLE_API_KEY": "your_gemini_api_key" } } } }

3. Restart Claude Desktop. 

Example usage: Open Claude and say, *"My security vendor just called and said they are 30 minutes late. Can you update the system and text the outdoor zone volunteers?"* Claude will instantly talk to your NPM package in the background, update the local database, and send the Twilio texts.

2. Standalone Cloud API (SSE mode)

If you want to build a custom web app or allow remote agents (like LangChain or custom scripts) to manage your event over the internet:

1. Rent a cloud server and run the package in HTTP mode:

   npx -y eventsorg-mcp --transport http

1. The server spins up and listens at your URL (e.g., https://your-server.com/sse).
2. Now, in your custom app, you can connect remotely using the official MCP SDK:

   import { Client } from "@modelcontextprotocol/sdk/client/index.js";

import { SSEClientTransport } from "@modelcontextprotocol/sdk/client/sse.js";

const transport = new SSEClientTransport("https://your-server.com/sse"); const mcpClient = new Client(); await mcpClient.connect(transport);

// Trigger tools remotely over the internet! await mcpClient.callTool({ name: "get_event_summary", arguments: {} });

*(Note: Ensure you have your environment variables set in your terminal or cloud dashboard before running the npx command).*

3. Developer Setup (Cloning)

If you want to modify the code or run the autonomous background agent (agent.ts):

1. Clone the repository and run npm install.
2. Copy .env.example to .env and fill in your keys.
3. Run npm run start (for the API) or npm run start:agent (for the autonomous worker).

Cloud Deployment

If you want to host this publicly so an AI can manage it 24/7 in the cloud, you can deploy this on any Node.js hosting provider (like Render, AWS, Railway, Heroku, or Fly.io):

1. Push this code to your own GitHub repo.
2. Connect your repository to your cloud provider and deploy it.
3. Paste your .env variables into your cloud provider's dashboard.
4. Ensure your provider gives you a public URL (e.g., https://my-events.onrender.com), and set this as the PUBLIC_URL environment variable.
5. In your Twilio console, point your phone number's webhook to https:// /webhook/sms.

*(Note: The npm start command will automatically seed the database with fake "sample event" data. If you are running a real event, remove tsx src/seed.ts from the start script in package.json!)*

---

🧪 Testing with the Live "Sample Event"

If you just want to test how the MCP protocol works without hosting a real event, I have deployed a public test instance of this server. Every time this test server boots, it completely wipes the database and fills it with a chaotic, half-broken music festival (vendors are 45 minutes late, zones are short-staffed).

You can build or connect your own MCP clients (like LangChain, custom apps, or AI code editors) to test it right now!

Live Test SSE Endpoint: https://eventsorg-mcp-production.up.railway.app/sse

Connect your AI to this URL and ask it: *"What is the status of my event?"* Watch as it identifies the issues and fires off physical text messages to try and save the festival!

---

🛠 Available MCP Tools

This MCP server exposes over 20+ tools to give your AI full operational control over the event. Here are the main highlights:

• 📱 Twilio SMS Orchestration: (send_sms, get_sms_history) Allows the AI to text staff directly. Inbound replies are automatically logged to the database via webhook.
• 📊 Google Sheets Dashboard: (sync_to_sheets) Pushes the real-time state of the database to a remote Google Sheet for human coordinators.
• 🗺️ Maps & Weather: (get_directions, get_weather) Provides the AI with live contextual data, such as walking times for reassigning volunteers or rain alerts for outdoor zones.
• 🔍 Core Operations: (get_event_summary, reassign_volunteer, cancel_vendor_slot, escalate_to_coordinator, move_zone_indoors)

*(Note: This is just a highlight! The server includes many more granular tools for managing incidents, checking in vendors, and querying exact database state.)*

---

📄 License

This project is licensed under the ISC License.