i’ve been using Hermes mostly through Telegram for almost 4 months.
which is sick - because i can talk to my agent from anywhere;
phone
laptop
walking around
whatever // wherever.
but after using the new Hermes Desktop app, the thing that clicked for me is pretty simple:
chat (TG, WhatsApp, Discord, etc...) is a great way to talk to your agent.
desktop is where you actually see the system around it.
that’s the part i think people are going to underrate.
Hermes Desktop gives your agent a home base.
you can see tools.
you can see skills.
you can see crons.
you can see sessions.
you can switch profiles.
you can see live agent runs.
you can see memory/context settings.
you can connect local or remote gateways.
if you’ve been operating through Telegram, Discord, WhatsApp, texts, or any other chat surface, this is a massive upgrade.
the main character here is visibility.
the killer feature: session continuity
the first thing that felt kind of magical:
i could start with my Hermes agent from my phone, then open Hermes Desktop and pick up the same session there.
same agent.
same sessions.
same skills.
same memory.
same config.
the conversation doesn’t disappear into a messaging app scrollback.
you can start something on Telegram, then move into Desktop when you need more control, more visibility, or just more screen space.
you can also go the other way: start something in Desktop, leave halfway through, then pick up that same session on your phone in Telegram.
that matters a lot if you actually use agents for work.
because a real agent session is rarely just “send one message, get one answer.”
it turns into:
- checking what happened
- continuing old context
- seeing what tools fired
- finding artifacts
- managing settings
- switching models
- watching context usage
- dealing with approvals
- jumping between different agents/profiles
that stuff is painful when everything is hidden behind chat.
Desktop makes it visible.
the app feels like an agent command center
when you open Hermes Desktop, the layout is familiar if you’ve used tools like Codex or Claude-style coding apps.
sessions/chats on the side.
main conversation in the middle.
controls/settings around the edges.
but the important thing is what it exposes.
you’re not just staring at a chat box.
you get a surface for the whole agent runtime.
some of the stuff that stood out to me:
1. agents view
there’s an Agents button that shows your current/live agent runs and sub-agents.
this is one of those features that sounds obvious until you’ve been running agents from chat apps.
in Telegram, i’m usually guessing what’s happening unless i ask.
in Desktop, i can actually see active runs.
that changes the feel from “i sent a message into the void” to “i’m watching my agent system operate.”
small difference, big deal.
2. ⌘K for everything
@imbabybrooklyn from the Hermes team posted a quick clip of one of the newer Desktop updates:
this is the kind of product detail that makes Desktop feel less like “a GUI for an agent” and more like an actual workspace for running one.
once Desktop exposes sessions, profiles, crons, tools, providers, gateways, model controls, context controls, memory budgets, and live agent views, clicking through sidebars gets old fast.
⌘K becomes the speed layer.
you hit the shortcut
type what you need
and jump straight there.
that matters because agent work is full of tiny context switches: opening old sessions, switching profiles, checking crons, changing models, inspecting tools, or jumping into settings without losing the thread.
a good command palette turns all of that into muscle memory.
small UI detail, but huge operator feel.
3. cron management
the Cron button is another one that immediately makes sense.
you can see the scheduled jobs your agent has.
you can pause them.
you can run them.
you can edit them.
fast.
if you’ve ever managed crons from config files or slash commands, this is just way nicer.
scheduled agent work becomes a visible thing you can manage instead of some invisible background behavior you hope is configured correctly.
4. switch agents with Manage Profiles
low-key sleeper feature:
top right icon → Manage Profiles.
this lets you switch between agent profiles.
i actually think this might end up being one of the best features in the app for power users.
because once you have more than one agent, the problem becomes less “can i talk to an agent?” and more:
which agent am i talking to?
what profile is active?
what tools does this one have?
what memory/context is attached?
what job is this profile supposed to do?
Desktop makes that way easier to reason about.
5. product vs technical tool-call display
you can choose how tool calls show up.
Product mode makes the experience feel more like ChatGPT’s web app. cleaner, less intimidating, easier to follow.
Technical mode shows more detail. more like Claude Code / coding-agent style traces.
that’s a really nice choice.
because sometimes you just want the product experience.
other times you want to see the guts.
if i’m casually using the agent, product mode is probably better.
if i’m debugging, testing, or trying to understand what happened, technical mode is what i want.
6. themes
there are 6 different theme families, with light and dark mode options for each.
so basically 12 different looks.
this is not the deepest feature in the world, but it does matter.
if this app becomes the place where you spend hours operating agents, the workspace should feel good.
and Hermes Desktop gives you enough visual control to make it feel like your setup instead of some default admin panel.
7. model + reasoning controls in the bottom trim
the bottom trim has model controls.
you can change the model.
you can change the thinking/reasoning amount.
you can do it from inside the desktop app.
that’s huge.
because model choice is one of the highest-leverage knobs when you’re using agents.
sometimes you want fast and cheap.
sometimes you want the expensive brain.
sometimes you want more reasoning.
sometimes you just need the thing to answer.
having that right inside the app makes the whole agent feel more controllable.
8. session timer + context window bar
this might be my favorite practical feature.
the bottom trim shows a session timer and a context window usage bar.
like an actual bar that fills up and tells you what percent of the context window you’ve used.
this is so helpful.
before, if you were using Telegram or Discord, you had to ask with something like /status or /usage.
now it’s just sitting there.
visible.
that one tiny UI element changes how you use long sessions.
you can tell when you’re getting close to the edge.
you can decide when to compress.
you can decide when to start fresh.
you can stop guessing why the agent is getting weird.
for people doing real multi-hour agent work, this is one of those “oh thank god” features.
9. approval timeout from Settings
this sounds silly until it saves you from rage.
Hermes has approval flows for things that need permission.
normally, those can timeout pretty fast.
like 60 seconds fast.
which is fine if you’re staring at the screen, but annoying if you’re using the agent from your phone, walking around, distracted, or doing anything remotely normal.
Desktop lets you change that approval timeout from Settings.
so you can set it to something realistic, like 5 minutes.
this is a perfect example of why Desktop matters.
it takes a thing that technically existed before and makes it way easier to operate.
less slash command hunting.
less config digging.
less “why did that request expire already?”
10. auto-compress context controls
you can also control automatic context compression.
turn it on.
turn it off.
change when it happens.
choose if it happens at all.
this is another serious agent-operator feature.
context compression can be awesome, but you want control over it.
sometimes you want the app to keep things moving.
sometimes you want to decide manually.
sometimes you’re doing something delicate and don’t want surprise summarization changing the shape of the session.
Desktop puts that control in front of you.
11. memory + user profile budgets
you can change the budget for your memory.md and user.md files.
this is nerdy, but it matters.
Hermes has memory and user profile context that gets pulled into sessions.
that context has a budget.
Desktop lets you tune it.
so if you care about how much durable memory or profile info your agent carries into a session, you can adjust that without spelunking through config.
again: visible controls for things that used to feel hidden.
12. providers, auth flows, and API keys
you can change/swap model providers from Settings.
you can trigger auth flows.
you can put in API keys.
you can manage provider setup from the app.
this sounds simple, but it’s actually hella convenient.
chat slash commands are cool when you already know exactly what you’re doing.
but for setup, auth, provider switching, and key management, a real settings screen is just better.
there’s less guessing.
less command memorization.
less “where does this config live again?”
13. local and remote gateways
you can wire Hermes Desktop into a local gateway or a remote gateway.
that means your agent can be running locally, or it can be a cloud/remote setup, and Desktop becomes the place you operate it from.
this is the part i showed in the quick tutorial.
Settings → Gateway → remote gateway
add the URL/session token
save + reconnect
old sessions load in
but the bigger point is that gateway setup is one chapter of the story.
the real unlock is what happens after it connects:
your agent has a visible workspace.
14. skills, tools, and MCP setup
skills, tools, and MCP setup are way easier to understand from Desktop.
you can see what your agent can do.
skills stop being some abstract folder of magic instructions.
tools stop being invisible capabilities buried in config.
MCP setup becomes something you can actually reason about from the app.
this is probably the best way to explain Hermes Desktop to someone who hasn’t used it yet:
it makes the agent’s capabilities legible.
you can see the stuff your agent has access to.
you can change the setup.
you can understand what’s wired in.
that is a very different experience from “message the bot and hope you remember what it can do.”
why this matters
agents are getting more powerful, but the control surfaces still feel weird.
a lot of us have been operating them from terminals, chat apps, config files, random slash commands, and half-remembered setup flows.
that works if you’re technical enough and stubborn enough.
but the whole thing gets way better when the agent has a real desktop home.
Hermes Desktop gives you:
- continuity across surfaces
- visibility into sessions
- live agent/sub-agent runs
- cron controls
- profile switching
- model + reasoning controls
- provider/auth setup
- context usage
- memory budget controls
- gateway setup
- skills/tools/MCP visibility
- product or technical tool-call views
all in one app.
that’s the masterclass here.
Hermes Desktop takes the parts of agent work that used to be hidden, scattered, or command-driven and gives them a place to live.
if you’re casually chatting with an agent, you might not need all of this yet.
if you’re actually operating agents for work, this is the kind of app that makes the whole system click.
and for me personally, it made the Hermes setup feel a lot less like “a bot i talk to from Telegram” and a lot more like an actual agent workspace.
go download it, connect your agent, and poke around the settings.
there is way more in there than i expected.
as Hermes Desktop from @NousResearch keeps shipping, i’ll keep updating you.
if you’re building with Hermes or agents in general, follow along.
-cobi
