Follow ZDNET: Add us as a preferred source on Google.

ZDNET's key takeaways

• Hermes is an AI desktop app with plenty of cool features.
• You can easily select models and check agent progress.
• Hermes learns as it goes and allows you to schedule queries.

AI isn't everyone's favorite topic these days, and I totally get it. I avoid the most heated issues by using AI only for research (and checking for accuracy as I go). I also use locally installed AI to avoid straining the electric grid and to retain my privacy.

In the grand scheme of things, AI isn't going away and does have valid uses, such as in the medical and scientific fields. Because of that capability, I cover AI because it's something people want to talk about, and I'm on a mission to ensure everyone knows that locally installed AI is the way to go. To that end, I test many apps that can be installed on machines in my local network.

Also: Want local vibe coding? This AI stack might replace Claude Code and Codex - for free

A relatively unknown app called Hermes came to my attention recently -- and when I noticed that it can be used with the open-source tool Ollama, you can bet I jumped on it.

Hermes isn't just a simple chat app where you can query it, though you can use it for that purpose. Instead, Hermes is an autonomous agent that lets you configure various tools, manage skills, view files, talk via voice, and keep several agent conversations organized in one place. Essentially, Hermes takes the chatbot concept and adds integrations, a terminal, a desktop app, messaging channels, memory, skills, scheduled jobs, and a learning loop.

Hermes is many things, but it never feels overwhelming. And where some AI desktop apps seem limited, Hermes practically shrugs off those limitations as if to say, "Anything you can do, I can do better… and faster."

Also: These top 30 AI agents deliver a mix of functions and autonomy

But what can Hermes do? To answer this question, you need to know what a Hermes Agent is.

What is a Hermes Agent?

The simple answer is a self-improving personal agent. But that definition doesn't say much, does it? 

Also: OpenAI's new image watermarks make it easier to spot AI fakes - here's how

As the official site says, "Hermes can remember what it learns, generate persistent skills, run scheduled automations, delegate to subagents, and use sandboxed backends like local, Docker, SSH, Singularity, and Modal."

In other words, Hermes can start at Point A (say a query from you), and then navigate through points B, C, D, E, F, G, and more. 

To really understand what an Hermes agent is, consider what it consists of:

• Memory: The information you give Hermes for future actions.
• Skills: Playbooks used to create reproducible actions.
• Soul: The constituent elements of an agent: voice, style, preferences, and default behavior.
• Crons: Scheduled automations that ensure the agent can behave proactively.
• Session recall: A searchable history the agent can use to remember previous conversations, links, decisions, and projects.

When you combine all the above elements, you get a reasoning loop that can read your messages, choose tools for the action, call up skills, update its memory, and then decide what action to take next.

Hermes involves many elements, which means it does have a bit of a learning curve. However, some features have elevated this app to the top of my list for desktop AI apps. 

Also: I tested ChatGPT Plus vs. Gemini Pro to see which is better - and if it's worth switching

Let me show you how I installed and used Hermes and explain why it's become one of my favorite AI tools.

Installing Hermes

There are two different ways to install Hermes: via the official installer or via Ollama. I tried both and found the Ollama option to be considerably easier. Ollama also allows me to use Hermes for free (in conjunction with the GPT-OSS model).

Also: US workers are the world's biggest AI skeptics - and it's not just about job loss

I installed Hermes on Pop!_OS Linux, but you can install it on MacOS and Windows as well. 

curl -fsSL https://ollama.com/install.sh | sh

curl -fsSL https://ollama.com/install.sh | sh

irm https://ollama.com/install.ps1 | iex

On Linux, after the installation, you have to start the Ollama service with the command:

sudo systemctl start ollama

You can now launch Hermes on any of the three operating systems using the command:

ollama launch hermes-desktop

The above command will open a window where you must select your model. The selection is limited, but we can select from more options once the GUI is running.

Adding a new model

The Hermes window should now be open, but you probably don't want to use the model you first selected.

1. Open Settings

Click the gear icon near the top-right corner of the window. From the resulting page, select the model you want to use, such as Google Gemini (OAuth). 

2. Onboard your model

Click the Set Up link below the dropdown and walk through the onboarding process. When the setup is completed, close the Settings window, and you should then be able to select the model you want to use.

Now's when things get cool. I ran the following query: "Build an app that allows me to keep an inventory of all of my vinyl albums."

Also: How AI agents will transform your customer service - despite 3 hurdles

Hermes went to work and began the process. The first thing that happened was that it asked me what kind of app I wanted to build. Interesting.

I opted for a Desktop app.

Next, I had to choose between Electron and a Python-based app. I went with Python (because I know it's installed on my machine).

Hermes continued in this fashion, prompting me for specific details for my app, such as the essential features I wanted. Hermes then went to work building my new app, while allowing me to watch its progress. 

At the bottom of the window, I was also able to see (in real time) how long the session had been running. This ability is important because it allows me to go back to any session and pick up where I left off -- all while knowing how much time I've spent.

In addition, Hermes indicated the time to run through each step.

Also: 96% of IT pros use AI now: Their top 7 agentic applications and biggest implementation roadblocks

Of course, it would probably have gone quicker had I used a model like Qwen Code, but I wanted to see how well Gemini would do. 

When the task was completed, it asked me if the app had opened. It hadn't. When I answered "No," it instructed me to install PyQT5. I was impressed that the tool knew my environment.

Sadly, the instructions it gave me to install PyQT5 were incorrect (it used Pip, but PyQT5 is an externally managed app). I knew the right command, which was:

sudo apt install python3-pyqt5

In the end, the app failed to run. Hermes attempted to help me, but I ran into my query quota.

Drats.

Why I dig Hermes

The reason why I would opt to use Hermes over similar apps boils down to some key factors:

• Ease of switching models.
• Watching it walk through each step of the process, while also tracking how much time each step took, as well as the session time.
• You can easily view the artifacts for each chat session, including any files Hermes created in the process.
• You can pin chats.
• You can set cron jobs, such as summarizing Slack threads each morning.
• You can interact with Hermes via speech.

The cool features go on and on, and the best way for you to see why this app grabbed my attention is to install it and use it.