How to master Hermes Desktop in under an hour

If you’ve been running Hermes through Telegram or WhatsApp, you’ve probably hit the same wall many users do: endless chat history, lost links, and no real way to manage multiple agents or long-running tasks. Hermes Desktop fixes that by turning Hermes into a proper control center instead of just another chat window.

This guide walks you through installing Hermes Desktop, connecting it to both cloud and local models, running Hermes 24/7 on a VPS, and using powerful features like profiles, artifacts, sub-agents, and cron blueprints to build a serious AI workflow.

Why switch from Telegram to Hermes Desktop?

Using Hermes through Telegram or WhatsApp is convenient, but it breaks down fast when you start doing real work. Multiple agents, long-running sessions, and files scattered across months of chat history quickly turn into chaos.

Hermes Desktop solves this by giving you:

• Structured sessions instead of one endless chat
• Pinned sessions for recurring workflows
• Profiles with different personalities, tools, and models
• A central artifacts view for every file, image, and link you’ve shared
• Built-in support for cron jobs, skills, and sub-agents

It’s still the same Hermes agent under the hood, but now you get a real interface designed for managing complex work, not just messaging.

Installing Hermes Desktop in a few clicks

Hermes used to require a terminal install with curl commands and manual config. Hermes Desktop replaces that with a simple installer.

To get started:

1. Go to the Hermes Desktop download page and grab the app for your OS.
2. Install it like any normal desktop app (e.g., drag to Applications on macOS).
3. Open Hermes Desktop and choose your LLM provider when prompted.

If you pick OpenAI, you just log in, confirm your account, and you’re ready to start a new session. Within a few minutes you can type “hi” and get a response from Hermes running locally, without touching the terminal.

Understanding Hermes, gateways, and LLM providers

Hermes has three main pieces you should understand:

• Control interface: This is Hermes Desktop, Telegram, WhatsApp, etc. It’s how you talk to Hermes.
• Hermes gateway: The actual Hermes process that runs your agents, skills, and automations. It can run locally or on a server.
• LLM provider: The model backend (OpenAI, Claude, local LLM via LM Studio, etc.) that Hermes calls to generate responses.

These three parts can all live on different machines. For example, you can:

• Use Hermes Desktop on your laptop
• Run the Hermes gateway on a VPS 24/7
• Connect that gateway to a local LLM on a separate machine, or to multiple cloud models

This separation is what makes Hermes so flexible—and Hermes Desktop makes configuring it much easier than editing config files.

Connecting Hermes Desktop to an LLM provider

Once Hermes Desktop is installed, you can configure models in the settings panel.

In the settings you can:

• Choose your primary model provider (e.g., OpenAI, Anthropic, etc.)
• Select specific models for general chat, coding, vision, or web extraction
• Add auxiliary models for special tasks (vision, web extract, etc.)
• Configure a fallback model so Hermes automatically switches if one provider goes down

You can also customize:

• Personality (e.g., more technical vs. more casual)
• Time zone
• Whether to show reasoning blocks
• Theme (dark, light, or custom themes like “Midnight”)
• Workspace directory and safety/approval settings
• Voice input and other advanced options

Previously, setting multiple providers and fallbacks required manual config. Now it’s all point-and-click inside the app.

Using local LLMs with LM Studio

If you want to avoid token costs and run models on your own hardware, Hermes Desktop integrates with LM Studio. This lets Hermes talk to a local LLM instead of (or in addition to) cloud models.

Step 1: Set up LM Studio

1. Download LM Studio from lmstudio.ai and install it.
2. Open LM Studio and download a model that fits your machine, ideally with at least a 64K context window so Hermes can use it effectively.
3. Go to the Developer tab in LM Studio and start the local server. Note the base URL (IP + port).

Step 2: Configure Hermes Desktop

In Hermes Desktop:

1. Go to Providers > API keys and find the LM Studio section.
2. Paste the LM Studio base URL as the base URL override.
3. In LM Studio, enable authentication in server settings, create a token, and copy it.
4. Paste that token into Hermes Desktop as the LM Studio API key.

One important step: Hermes expects a 64K context window. In LM Studio’s Local servers tab, increase the context slider to at least 64,000 tokens, press Load, and select that reloaded model.

After that, you can switch to your local model from the bottom-right model selector in Hermes Desktop and chat as usual—no per-token costs, just your hardware and electricity.

Running Hermes 24/7 on a VPS

If you only run Hermes locally, it stops when your computer sleeps or shuts down. That means no cron jobs, no background monitoring, and no always-on automations.

To keep Hermes running 24/7, you can install it on a VPS. The general flow is:

1. Sign up with a VPS provider and choose a plan (something like an 8 GB RAM plan is often a good starting point).
2. Choose Ubuntu LTS as the operating system for stability and long-term support.
3. In the VPS application marketplace, select the Hermes app image and deploy it.
4. Remember the admin password you set during deployment—you’ll need it for the Hermes web dashboard.

Once Hermes is running on the VPS, connect it to an LLM provider:

1. Open the VPS terminal from your hosting dashboard.
2. Navigate to the Hermes Docker folder (e.g., cd /docker).
3. Run the Hermes setup command (e.g., docker compose exec -it hermes setup).
4. Choose your provider (like OpenAI), paste your API key and base URL, and pick a default model.

After setup, you can confirm it’s working by opening the Hermes web UI, starting a chat, and sending a quick message.

Using the remote gateway with Tailscale

Hermes Desktop can act as a remote control for a Hermes gateway running elsewhere—like your VPS or a mini PC at home. This is what the remote gateway feature is for.

Why use a remote gateway?

• Your Hermes runs 24/7 on a server.
• You can connect from multiple devices (desktop, Telegram, WhatsApp) to the same shared Hermes.
• You keep heavy workloads off your laptop.

Step 1: Set up Tailscale on your desktop

1. Go to tailscale.com and create an account.
2. Download and install Tailscale on your computer (e.g., macOS standalone variant).
3. Sign in and make sure your machine shows as connected in the Tailscale dashboard.

Step 2: Install Tailscale on your VPS

1. Open your VPS terminal from your hosting control panel.
2. Follow the Tailscale Linux instructions—typically a single curl command to install.
3. Run sudo tailscale up and authenticate when prompted.
4. Confirm your VPS appears as a connected Linux device in the Tailscale dashboard.

Step 3: Connect Hermes Desktop to the remote gateway

1. On the VPS, find the Hermes gateway container and note the exposed port (e.g., 0.0.0.0:32769).
2. In Tailscale, copy the VPS’s Tailscale IP (e.g., 100.122.x.x).
3. In Hermes Desktop, go to Settings > Gateway and switch from Local to Remote.
4. Enter the remote URL as http://<tailscale-ip>:<port>, then save and connect.
5. Log in with the same username and password you use for the Hermes web dashboard (default user is often hermes plus your saved admin password).

Now your Hermes Desktop is just an interface, while the real Hermes gateway runs on your VPS 24/7. You can verify it by sending a message from the desktop and seeing the same session appear in the web UI on the server.

Boosting search with Tavily and other tools

Hermes can use your LLM provider for web search, but that can be token-heavy and noisy. A dedicated search API like Tavily gives you cleaner results and often saves money.

To improve search:

1. Sign up at tavily.com and grab your API key.
2. In Hermes Desktop settings, paste your Tavily key into the web search tools section.

Hermes also supports other search and crawling tools like Firecrawl, Brave Search, and more. Any of these can be plugged in from the same settings area to give Hermes richer, more targeted web access.

Working with sessions, voice, and model switching

Hermes Desktop organizes your work around sessions. Each session is like a dedicated workspace for a task or project, instead of one giant chat log.

Key session features:

• New session: Start a fresh conversation for a new task.
• Pinned sessions: Pin important or recurring workflows (e.g., daily social media planning) so they don’t get buried.
• Grouped sessions: Optionally group sessions into workspaces or projects.

Hermes Desktop also supports:

• Voice dictation: Click the mic, speak your request, and Hermes transcribes it locally. This is much faster than typing for long prompts.
• Conversation mode: Hermes replies with audio, so you can have a back-and-forth voice conversation without constantly pressing enter.
• Quick model switching: A model selector in the bottom bar lets you swap between models (e.g., GPT-5.5, Claude, code-focused models) with a single click.

Managing multiple agents with profiles

Profiles are one of the most powerful additions to Hermes Desktop. A profile is a complete configuration that includes:

• Its own souls.md (personality and behavior instructions)
• Its own default models and tools
• Its own separate session history

Instead of one Hermes that does everything, you can create specialized profiles:

• Default: General-purpose assistant for everyday tasks.
• Archiver: A “link curator and knowledge organizer” that saves links, GitHub repos, articles, and posts into your Obsidian vault with tags and summaries.
• Researcher: A research briefing agent that turns messy questions and saved links into structured research notes.

Each profile has its own souls.md file describing how it should behave. For example, the Archiver profile’s soul tells Hermes to:

• Treat the user as someone sending links all day
• Store them in a specific Obsidian folder
• Add type labels and tags
• Generate a short summary for each item

The result is a clean, persistent second brain: Hermes ingests your links and outputs organized Obsidian notes.

You can also create model-specific profiles, like:

• Coder profile that always uses a code-optimized model
• Writer profile that defaults to a model tuned for long-form writing

This way you don’t have to manually switch models every time—you just switch profiles.

Because profiles live on the same Hermes gateway, you can even share one Hermes instance with another person (e.g., a partner or teammate) by giving them Hermes Desktop and a dedicated profile, though note they can still see the shared history unless you isolate instances.

Artifacts: a searchable memory for files, images, and links

One of the biggest pain points with chat-based agents is finding that one file, screenshot, or link you shared weeks ago. In Telegram or WhatsApp, that usually means endless scrolling.

Hermes Desktop introduces artifacts—a centralized view of all non-text content you’ve given Hermes:

• Images
• Files
• Links
• Generated assets like HTML files

From the artifacts panel you can:

• Quickly search for past resources
• Filter by type
• Open live previews (for example, HTML pages Hermes generated)

This is especially useful if you use Hermes as a research assistant or content production hub, where you constantly pass it documents, screenshots, and URLs.

Sub-agents and live artifact previews

Hermes Desktop also introduces a more advanced concept: sub-agents. These let Hermes run multiple tasks in parallel without blocking your main chat.

Here’s how it works:

• A normal agent session can only run one task at a time, tying up the chat window.
• When Hermes detects a complex job that can be split, it can spawn sub-agents to handle parts of the work in the background.
• You can keep chatting in the main session while sub-agents run tasks like code analysis, data fetching, or file generation.

Hermes Desktop shows:

• How many agents and sub-agents are currently running
• A breakdown of tasks into subtasks
• Status indicators as those tasks complete

When Hermes generates files (like a self-contained HTML dashboard of World Cup results), you can open a live artifact preview directly in the app. Instead of scrolling through raw HTML in chat, you see the rendered page with interactive elements like group tables and charts.

This combination—sub-agents plus live previews—makes Hermes Desktop far more capable than a simple chat interface for serious development and automation work. If you’re interested in deeper model strategy for Hermes, you may also find this guide to top AI models for Hermes agents useful.

Cron jobs and automation blueprints

Cron jobs let Hermes run tasks on a schedule, even while you sleep. Typical use cases include:

• Monitoring GitHub repos for changes
• Checking competitor sites for updates
• Pulling daily news or research summaries
• Sending regular reports to Telegram or email

Hermes Desktop offers two ways to create cron jobs:

1. Automation blueprints: Pre-built recipes you can copy and paste.
2. Visual cron editor: A GUI where you define name, prompt, frequency, and delivery destination.

Using automation blueprints

On the Hermes website you’ll find an Automation blueprints section. For example, a “competitive update scout” blueprint can monitor multiple GitHub repos (like Hermes, Claude Code, or Codex) and summarize what changed.

To use a blueprint:

1. Copy the blueprint text.
2. Paste it into a Hermes Desktop session.
3. Provide any required inputs (e.g., repo URLs).
4. Hermes will install any needed skills and set up the cron job.

The new cron appears in the Cron panel, where you can:

• See its schedule
• Trigger it manually
• Edit or delete it

If you prefer manual setup, click Cron in the sidebar, hit New, and define your own schedule and prompt from scratch.

Skills, tools, and the Skills Hub

Skills extend Hermes with reusable capabilities so you don’t have to prompt from zero every time. They’re like packaged workflows or tools that Hermes can call when needed.

In Hermes Desktop you can:

• See a list of installed skills
• Toggle skills on or off with a click
• Manage configurations for each skill

Hermes also offers a Skills Hub (at /doc/skills on the Hermes site) where you can browse community and official skills. Installing them is simple:

1. Find a skill you like in the Skills Hub.
2. Copy the install command or snippet.
3. Paste it into a Hermes Desktop session.
4. Hermes installs the skill and it becomes available in your environment.

With the right mix of skills and cron jobs, Hermes can act more like a full automation platform than a simple chatbot. If you enjoy building AI-powered workflows on desktop, you might also like the workflow-focused guide to Claude Code for desktop.

Connecting messaging apps: Telegram, WhatsApp, and more

Even if you switch to Hermes Desktop as your main control center, you can still connect messaging apps for on-the-go access. Hermes Desktop makes this configuration much easier.

Supported channels include:

• Telegram
• WhatsApp (including WhatsApp Enterprise)
• WeChat and others

For example, to connect Telegram:

1. In Hermes Desktop, go to the Messaging section.
2. Paste your Telegram bot token from BotFather into the Telegram configuration.
3. Optionally set allowed user IDs for security.
4. Restart the Hermes gateway from your VPS control panel so the new config takes effect.

Once it restarts, your Telegram bot will be connected to the same Hermes gateway your desktop uses. In Hermes Desktop you’ll see a Telegram icon and can view sessions that originated from Telegram separately from desktop sessions. Multiple people can message the same bot, all routed through your central Hermes instance.

Pros, limitations, and why Hermes Desktop is worth it

Hermes Desktop addresses two major pain points of chat-only control:

• Organization: No more drowning in massive Telegram threads. Sessions, profiles, pinned chats, artifacts, and workgroups make it easy to find what you need.
• Setup complexity: Installing Hermes, wiring models, setting up a VPS, and connecting messaging apps used to be intimidating. The desktop app turns most of this into a guided, click-based flow.

There are still some trade-offs:

• It’s a desktop app, so you need Telegram/WhatsApp for full mobile convenience.
• The app is relatively new, so there are rough edges—like file/folder views still reflecting local files even when you’re connected to a remote gateway.
• Features are evolving quickly, which means frequent updates (a plus for improvements, but something to be aware of).

Despite that, Hermes Desktop is currently the best way to get started with Hermes if you’re serious about agents, automations, and multi-model workflows. Install it, connect a model, set up a remote gateway when you’re ready, and then lean into the advanced features—profiles, artifacts, sub-agents, cron blueprints, and skills—to really unlock what Hermes can do.