Deploying and Extending Hermes Agent on DigitalOcean

In the year 2026, AI agents have begun executing tasks directly for users, marking a shift from merely providing guidance. Among these advancements is the Hermes agent, an open-source, self-improving AI developed to enhance user interaction by learning from experience, refining skills, and building a comprehensive model of the user over time.

This guide details the deployment of Hermes on a DigitalOcean Droplet, connecting it to Telegram, and extending its capabilities with a custom skill. As an example, we will create a grocery tracking agent that monitors daily consumption, alerts users when stock is low, and automatically places orders through a grocery delivery service.

Key Takeaways

• Hosting Hermes: Deploy Hermes on a DigitalOcean Droplet using Ubuntu 24.04 with a minimum of 2 vCPUs and 4 GB RAM. Installation is streamlined via the official script, followed by running hermes setup to configure your LLM provider.
• Telegram Integration: Utilize hermes gateway setup to connect Hermes to Telegram, maintaining accessibility even after disconnecting from SSH.
• Skill and MCP Configuration: Skills are defined in Markdown files located in ~/.hermes/skills/. MCP servers are configured in Hermes to expose features like grocery search and cart actions.
• HTTP MCP and OAuth: For headless server authentication, use an SSH tunnel from your local browser to the Droplet's loopback port.

Prerequisites

To begin, ensure you have:

• A DigitalOcean account and a Droplet running Ubuntu 24.04 with at least 2 vCPUs and 4 GB RAM.
• A Telegram account.
• An API key from a supported LLM provider (e.g., Anthropic, OpenAI).

Building the Hermes Architecture

Hermes is versatile, connecting with any tool supporting the Model Context Protocol (MCP), and it communicates via platforms like Telegram, WhatsApp, Discord, and Slack.

Step 1: Creating the Droplet

Log into your DigitalOcean account and create a new Droplet:

• Region: Choose the nearest location.
• Image: Select Ubuntu 24.04 LTS.
• Plan: The basic plan with at least 2 vCPUs and 4 GB RAM.
• Authentication: Use an SSH Key.

SSH into the Droplet and update the system:

ssh root@YOUR_DROPLET_IP
apt update && apt upgrade -y

Step 2: Installing Hermes Agent

Use the installation script to set up Hermes:

curl -fsSL https://raw.githubusercontent.com/NousResearch/hermes-agent/main/scripts/install.sh | bash
source ~/.bashrc
hermes setup

Choose your LLM provider and enter the API key.

Step 3: Connecting Hermes to Telegram

Set up the gateway for Telegram:

hermes gateway setup

Follow the steps to create a bot with Telegram’s BotFather, then start the gateway:

hermes gateway start
systemctl enable hermes-gateway
systemctl start hermes-gateway

Send a message to your bot to confirm the connection.

Step 4: Understanding Skills and MCP Servers

Skills are task-specific instructions in Markdown files under ~/.hermes/skills/. MCP Servers allow Hermes to access external services, enabling structured API communication.

Step 5: Building a Grocery Tracking Automation

Create a skill to automate grocery tracking using MCP services like Swiggy Instamart:

• Create a Skill File:

mkdir -p ~/.hermes/skills/grocery
nano ~/.hermes/skills/grocery/grocery_tracker.md

Paste and customize the following:

# Grocery Auto-Order Skill

## My Daily Grocery List

groceries:
  - name: "Milk"
    unit: "ml"
    daily_consumption: 500
    reorder_quantity: 2000
    alert_threshold_days: 3
    search_query: "fresh milk 1 litre"
  # Additional items...

• Create Inventory File:

nano ~/.hermes/grocery_inventory.json

{
  "last_updated": "2026-05-07",
  "delivery_address": "YOUR FULL ADDRESS HERE",
  "items": {
    "Milk": { "quantity": 1000, "unit": "ml" }
    // Additional items...
  }
}

• Connect MCP Server:

hermes config edit

Add the MCP server configuration and verify:

mcp_servers:
  swiggy-instamart:
    url: "https://mcp.swiggy.com/im"
    auth: oauth

Step 6: Authenticating with Your MCP Server

Log into the MCP server using an SSH tunnel:

hermes mcp login swiggy-instamart
ssh -i ~/.ssh/id_ed25519 -L 45123:127.0.0.1:45123 root@YOUR_DROPLET_IP

Open the provided URL in your browser to complete the authentication.

Step 7: Testing the Full Flow

Test the grocery tracking:

Check my groceries and tell me what's running low.

Reply with YES to place orders through the connected grocery service.

Conclusion

You've successfully deployed an AI agent on DigitalOcean that operates continuously. Hermes integrates with various messaging platforms and can be extended to manage a multitude of tasks through skills and MCP servers. This guide offers a foundation for further automation projects, such as bill reminders, health tracking, or home automation.