For a while, I kept recommending Hermes Agent to knowledge workers in this newsletter. The honest caveat every time was: you need to be comfortable in a terminal.

Install via a curl command. Edit YAML files to configure it. Run the gateway in a separate terminal window. Connect to it via Telegram or a text-based dashboard. Want it always on? Host it on a Mac mini or a VPS and SSH into it.

Powerful? Absolutely. Approachable? May not for all.

One Topic: Hermes Desktop App Personal AI Agents Just Got Accessible

The Surface Release

Nous Research shipped Hermes Agent v0.16.0, called “The Surface Release.” The name is deliberate. The agent now meets you on the surfaces where you actually work.

The headline: a native desktop app, available for Mac, Windows, and Linux.

Not a web app you open in a browser. Not a terminal wrapper with a slightly friendlier interface. A real desktop application. You download it, drag it into Applications on Mac or run the installer on Windows, and you are done. No Python setup. No YAML. No terminal window staying open in the background.

This is the version that turns Hermes from a developer tool into something a marketing manager, a product lead, or a consultant can actually adopt.

What changed compared to before

Previously, the full Hermes workflow looked like this. Install via terminal. Configure API keys in a text file. Launch the gateway in one window. Open a TUI or browser dashboard in another. Communicate with the agent via Telegram or a separate chat interface.

Each piece worked, but you were always managing pieces.

Now the Desktop app is the control room. Everything lives in one place.

The left sidebar shows your running sessions and pinned workspaces. The centre is a clean chat interface with streaming output so you can watch the agent think and act in real time. A side panel shows the active project directory, so you can see the agent reading, writing, and editing your files as it works. A right-side rail renders web pages, documents, and images directly alongside the conversation.

And all of it: API keys, memory settings, messaging channels like Telegram or Slack, scheduled tasks, skills, profiles, is configurable through a proper visual settings panel. No file editing. No terminal commands.

What you can do with it

For readers who are new to this newsletter, here is a quick grounding. Hermes is not a chatbot. It is an agent with persistent memory. Every conversation is stored and searchable. When it solves a complex task, it saves the reasoning pattern as a skill file so it can reuse it next time without starting from scratch.

The Desktop app exposes six core capabilities through that clean interface.

1. Connect everywhere. Hermes runs across Telegram, Slack, Discord, WhatsApp, Signal, email, and more. One agent, one memory, across every surface. Start a task at your desk, get updates on your phone via Telegram. The context follows you.
2. Remember your work. Every conversation is stored in a searchable database. The agent builds a long-term profile of your preferences, your projects, your working style. It auto-generates skills from patterns it notices. You do a thing twice, it knows how to do it the third time without being asked.
3. Schedule recurring tasks. Natural language is all it takes. Tell it to “send me a briefing of today’s meetings and top emails every weekday at 7:45 AM.” It sets it up, runs it unattended, delivers it to your inbox or Telegram. No code. No cron job configuration.
4. Delegate complex work. For big tasks, Hermes spawns isolated subagents, each with its own workspace and terminal. The main agent coordinates, the subagents do the heavy lifting in parallel. This keeps things fast without losing context.
5. Browse and research. It has over 40 built-in tools: web search, browser automation, vision, image generation, text-to-speech. It can research a topic, pull data from web pages, and generate multimedia files directly in its workspace.
6. Run experiments safely. Because it executes real code and terminal commands, it uses isolated sandboxing. You can choose Docker containers, remote servers via SSH, or serverless platforms. The host machine stays protected.

How to actually start

Download from official Hermes website. On Mac, it is a DMG. On Windows, a standard installer that handles all dependencies automatically, including Python, without touching anything else on your machine. First launch takes 10 to 15 minutes to bootstrap, then you are in.

From there, the Quick Setup path connects you to a model through Nous Portal in seconds. No manual API key configuration required for a first run.

Once inside, the shift worth making is this: treat the setup like onboarding a new team member, not testing a new app. Write a short user profile. What your role is, what you want delegated, how you like things communicated. Create two or three specialist profiles: one for inbox and operations, one for research, one for content. Set one scheduled briefing. That is enough to start seeing the difference.

The barrier that kept this tool in the hands of developers for months is gone. The question now is a simpler one: what repeating work are you still doing yourself that an agent could handle?

Enjoyed this? Forward it to someone still running AI from a browser tab.

Interested in travel or photography, read last week’s LensLetter newsletter about astro photography gears (but not expensive)

Read last week’s JustDraft about 7 GitHub Tools Your AI Agent Needs

Two Quotes to Inspire

The professionals who win the next decade won’t be the ones who use the most AI tools. They’ll be the ones who taught one agent to think like them.

Delegation is not a skill you practice once. It’s a system you build so that the work moves without you having to push it every time.

One Prompt to Steal

The Weekly Brain Dump Prompt

Use this every Friday before you close your laptop. Paste it into Claude, ChatGPT, or any AI you use regularly.

 Here is everything on my mind from this week. I'm going to do a 
brain dump. Your job is to:

1. Identify the 3 things I actually moved forward this week
2. List the 3 things I said I would do but didn't
3. Flag anything that sounds like it's stuck and ask me one question about it
4. Suggest one thing I should do first on Monday morning based on what I've shared

Ready? Here goes: [write freely for 2-3 minutes without editing]

Why it works: most weekly reviews ask you to fill in a structured template, which takes energy. This prompt takes your unfiltered thinking and structures it for you. The one Monday question is the most useful part. It forces a concrete next action instead of a vague intention.

Takes 5 minutes. Saves you from starting Monday wondering where to begin.