2 min read

How I Installed AgentForge with Docker Compose in 15 Minutes

How I Installed AgentForge with Docker Compose in 15 Minutes

Hook

Last Tuesday, a client asked me to set up an AI agent that could handle Telegram support tickets overnight. The catch: it had to run on their own server, not on someone else's cloud. They wanted control, not a subscription.

Background

I run a small DevOps shop called 1it.pro. We build internal tools for teams that need automation without the vendor lock-in. Hermes AI agents sounded right for this client, but I had never deployed AgentForge before.

The client runs a logistics company with 12 people. The support channel was a Telegram group where messages piled up after 6 PM.

Problem

The client had no after-hours coverage. Human operators logged off and the chat went silent. Important customer requests sat there until the next morning. The delay caused frustration and, once, a lost contract because nobody replied to a pricing question for 11 hours.

I needed a solution I could deploy fast, with full control over the data, and a clean way to configure the agent without SSH tutorials.

Why it matters

For a small team, hiring night shift support is not an option. Every unanswered message is a missed sale or an irritated customer. The cost of doing nothing was tangible: lost revenue and a reputation for being slow.

Outsourcing to a bot platform meant exposing customer messages to a third party. The client's privacy policy forbade that. Running locally was the only option.

Solution

I installed AgentForge on a $20 Ubuntu server. The whole thing took one Docker Compose file and a few environment variables. Here is what the stack looked like:

  • 🚀 A Docker Compose file with three services: agent, web UI, and a Redis queue
  • 🔧 Environment variables for the Telegram bot token, agent name, and system prompt
  • 🖥️ A web UI that runs on port 8080 for configuring the agent's behavior

The compose file is available in the AgentForge repository. I copied it, filled in my Telegram token, and ran docker compose up -d. That was it. The agent was online in under a minute.

I configured the agent through the web UI to respond to support questions, set a custom system prompt that matched the company's tone, and bound it to the Telegram group.

If you control the compose file and the environment variables, you control the agent. No external API calls. No data leaving your server.

Result

The agent has been running for six weeks without a single restart. It answers routine questions about shipping status, order changes, and pricing. Human operators get a transcript of every conversation the next morning and can follow up on anything the agent could not resolve.

The client's response time dropped from 11 hours to under 1 minute during off-hours. The Telegram group feels alive at 2 AM, and first-contact resolution went up by 40 percent.

I delivered the full setup in one afternoon, and I have reused the same compose stack for two other clients since then. That is the practical outcome of a clean Docker Compose install: one file, one server, and an agent that works.

AgentForge

AI agents for real work. If you want to see how this works in practice, visit our product site and blog.

Share
Serhii

Serhii