The argument in one line.
Hermes Agent 0.17 turns background delegation, multi-channel access, and self-improvement from manual opt-in features into automatic defaults, making it function less like a chatbot and more like a persistent ambient operating layer.
Read if. Skip if.
READ IF YOU ARE…
- You are already running Hermes agent and want to know exactly what changed in 0.17 and whether to upgrade immediately.
- You want to reach your AI agent from your iPhone without switching apps or opening a laptop.
- You are running long multi-hour research tasks in Hermes and need to keep chatting with it while it works.
- You are exploring AI-assisted game development and want to know if Unreal Engine 5.8 MCP is worth trying.
- You want to run multiple specialized Hermes agents with different models and skills without editing config files by hand.
SKIP IF…
- You do not use Hermes agent — this video covers zero Claude, GPT, or general-purpose AI assistant content.
- You are looking for a beginner introduction to AI agents; this assumes you already have Hermes running.
- You are on Android — iMessage integration is iPhone-only.
The full version, fast.
Hermes Agent 0.17 ships eight meaningful upgrades rather than cosmetic ones. The two biggest: background agents are now automatic for any sufficiently complex prompt (you can keep chatting while sub-agents work), and native iMessage support via a free Photon account means your phone becomes a full control terminal. The desktop app gets detachable windows, a model selector, and a built-in terminal. A new dashboard GUI handles profile creation and skill browsing. The agent also now self-patches its skill files during normal use without prompting, and Telegram gets full rich-text formatting including tables.
Free for members
Chat with this breakdown — free.
Sign in and you get 23 free chat messages on us — ask for the hook, quote a framework, find the exact transcript moment, generate a markdown action plan. Bring your own key when you want unlimited.
Create a free account →Where the time goes.
00:00 – 00:26
01 · Intro / 8-update preview
Host opens with competitive claim and shows the 8-item list card on screen.
00:26 – 02:51
02 · Native iMessage via Photon
Demo of texting Hermes from iPhone through Photon; investor report on Micron returned in iMessage. iMessage vs Telegram vs desktop decision matrix.
02:51 – 05:12
03 · Background agents (auto-enabled)
Background agents now fire automatically on complex prompts. New sub-agent tree UI shows 5 agents / 28 tool calls running live. Can interrupt and add tasks mid-run.
05:12 – 06:23
04 · Unreal Engine 5.8 MCP
UE5.8 ships first-class MCP support. Hermes can now build 3D games. Positioned as upgrade path from Three.js.
06:23 – 07:56
05 · Desktop app improvements
Detachable chat windows, model/thinking-level selector, live sub-agent pane, built-in terminal.
07:56 – 09:42
06 · Profile builder
GUI wizard in Hermes dashboard to create named profiles with specific models and skills. Replaces manual config-file editing.
09:42 – 10:50
07 · Skills hub
Browseable marketplace of hundreds of skills. Per-skill security scan. Creator tip: fork skill.md and let agent build a custom version.
10:50 – 12:02
08 · Smarter self-improvement
Agent now auto-patches its own skills during normal sessions. Performance improves organically.
12:02 – 13:05
09 · Telegram rich formatting
Tables, bold, streaming, lists now render natively. Demo of stock portfolio table.
13:05 – 13:38
10 · Outro / how to implement
Host reassures viewers they can ask Hermes itself to set up any feature from the update.
Lines worth screenshotting.
- Hermes background agents are now on by default — any complex prompt automatically spawns sub-agents so you can keep chatting while the work runs.
- A single free Photon account gives your iPhone iMessage thread full bidirectional access to a Hermes agent running on your desktop.
- Five background sub-agents doing 28 parallel tool calls is now a normal Hermes session, not a custom setup.
- Unreal Engine 5.8 is the first version of UE to ship MCP support, making AI-driven 3D game development possible without a browser engine.
- The Hermes skills hub lets you browse hundreds of community skills, but the smarter move is to fork the skill.md and have your agent build a custom version tuned for your workflow.
- Hermes now self-patches its own skill files during ordinary sessions — performance improves passively without any user action.
- The Hermes dashboard profile builder replaces config-file editing with a GUI wizard: pick model, add skills, name it, done.
- Telegram now renders full markdown tables from Hermes — you can ask for structured financial or research data and get a formatted table back.
- The desktop app now supports detached chat windows, so you can run separate Hermes agents side by side in different windows.
- iMessage is for quick go-anywhere queries; Telegram is for complex threaded deep-work sessions on the go — they serve different cognitive modes.
Eight habits that make an AI agent feel ambient.
WHAT TO LEARN
The gap between a chatbot you interrogate and an agent that works alongside you closes when you shift from single-turn commands to delegated, parallel, always-on workflows.
- Route quick on-the-go queries through the messaging app already open on your phone — friction at access time determines how often you actually use the tool.
- For long research or multi-step tasks, structure your prompt to be complex enough that the agent will parallelize automatically rather than blocking your conversation.
- Watching a sub-agent tree in real time (which models, which tool calls, which tasks) is the fastest way to understand where your workflow is bottlenecking.
- A browseable skills marketplace is only as useful as your willingness to fork and customize — installing a skill as-is means you inherit someone else assumptions about your workflow.
- Self-patching memory means an agent that uses a new tool today will be meaningfully better at it tomorrow without any manual intervention — run it more, not less, when it is struggling.
- Separating agents by model and purpose (one for coding, one for memory, one for research) is cheaper and more reliable than trying to prompt a single general agent into multiple specialized modes.
- Rich formatting in a messaging interface shifts the agent from a search-box replacement into a true information terminal — ask for tables and structured data, not just prose answers.
Terms worth knowing.
- Hermes Agent
- An open-source local AI agent framework by Nous Research that runs on your computer, manages skills and memory, and can control your system, browse the web, and interact with external services.
- Background agents
- Sub-processes that Hermes spawns to handle long-running tasks in parallel, freeing the main chat thread so you can continue sending messages while work is in progress.
- MCP (Model Context Protocol)
- An open protocol that lets AI agents connect to external applications and APIs as structured tools; Unreal Engine 5.8 is the first version of UE to support it natively.
- Photon
- A free service that assigns a phone number you can text, routing messages directly into a connected Hermes agent on your desktop.
- Skills (Hermes)
- Markdown instruction files (.skill.md) that teach Hermes how to use a specific tool or follow a specific workflow; the agent can read, write, and patch its own skill files.
- Profiles (Hermes)
- Named Hermes instances that each run a different model and skill set, enabling a multi-agent setup where separate profiles handle coding, memory management, research, and so on.
- Sub-agent tree
- A new UI panel in Hermes 0.17 that shows every active background sub-agent, what task it is working on, and which tool calls it is executing in real time.
Things they pointed at.
02:37toolPhoton
05:12toolUnreal Engine 5.8
07:56toolHermes dashboard
Lines you could clip.
03:17
“Now it can do work in the background, and you can keep chatting with it.”
03:35
“I have five agents doing 28 different tool calls right now to do this research.”
10:20
“I take the skill dot m d file, I give it to my agent, I say make your own version.”
13:05
“You have super intelligence at your fingertips.”
Word for word.
Read-along
Don't just watch it. Burn it in.
See every word as it's spoken — crank it to 2× and still catch all of it. The same dual-channel trick behind Amazon's Kindle + Audible.
analogy
00:00Hermes agent just had its biggest update ever. It is safe to say it is now way better than OpenClaw. In this video, I will cover eight of the big massive changes you need to know about that you need to start implementing today.
00:15Let's not waste any time. Let's get straight into it. There are eight big changes we're gonna talk about here.
00:20I'm kinda foreshadowing the first two here, a little bit of showmanship for you. But let's get right into the first one.
00:25I'm gonna show you how to set all this up, how to use it, everything. So don't worry about that. The first big update is native iMessage support.
00:34You now can chat with your Hermes agent directly through iMessage on your phone. So maybe I'm out on the go. I wanna check the Micron price.
00:42I wanna see what moved the price sale. Say, hey. Tell me the Micron price and why it moved today.
00:50I'm gonna hit send on that. This is directly in iMessage. As you can see, it has all the iMessage functionality.
00:56You can see that it's typing. All the rich iMessage features are in there.
01:01And as you can see, it it texts just like I do. Sends message to message to message. But what's great is a few things here.
01:08One, it's in the app you use all the time. Right? Everyone uses iMessage the most.
01:12Two, it's directly connected to Hermes on your computer. So if I go on here and I say, hey. Pop open a browser and do this.
01:19Go on my computer and move this file around. Go on my DGX Spark and install this local model. It'll do that for me.
01:26And as you can see, boom, a full nicely formatted investor report on Micron right there. Shout out to me. Invested in February.
01:34Should you be using this over Telegram, though? That is the question today or the desktop app.
01:40Here is how iMessage fits into your setup. I am now using iMessage when I'm on the go.
01:47It's the app that's easiest to boot up. It's the app that's easiest to chat with. As you can see here, I have Hermes actually pinned to the top of my iMessages so I can chat with it easily.
01:56If you're on your iPhone and on the go, this is the absolute easiest, quickest, simplest way to message your Hermes agent. So if on the go, I'm using iMessage. If I'm at my computer, I'm using the desktop app.
02:07I'm only using Telegram on the go now if I want to take advantage of threading. Right? When you're in Telegram, you can have many different subjects.
02:15If I'm doing complex deep work on the go, I'll go telegram. But most of the time, I'm on the go. I'm just firing off quick messages, and that's where iMessage is perfect for.
02:24This is all run through a service called Photon. It's free to use. Basically, gives you a phone number for you to text that turns into your Hermes agent.
02:33All you need to do is say this, hook Hermes up to iMessage using Photon according to the best practices from the new update. You hit enter on that.
02:41You'll be good to go. Hermes will hook iMessage up for you and create the account and give you the phone number. You might have to click on a link to log in.
02:48Easy peasy. You're good to go. You got it hooked up to your iMessage.
02:51The next big update, we're going through eight of them here. A little hint, a little preview. Background agents.
02:57So previously in Hermes, background agents needed to be turned on manually. You needed to flip a flag.
03:03And what background agents are is Hermes agent ability to spin up sub agents that do work in the background so you can keep talking to Hermes while the sub agents do work. This is very, very convenient.
03:17Typically, when you give Hermes a long task to do, it just sits there silently and works. You can't message it.
03:24You have to wait sometimes hours for it to work. Now it can do work in the background, and you can keep chatting with it. And now it's done automatically.
03:31No flags to flip. All you need to do is give a sufficiently complex enough prompt, and it will automatically use the background agent.
03:40So I'm giving it this prompt around writing, uh, complex research reports for a bunch of different AI companies I'm investing in. If you wanna take a screenshot here and take a look at these, these are all companies I've invested in. It's gonna load all the relevant skills, which is awesome, which is a big strength of Hermes agent.
03:57And what you can see here is a few things. First, you can see the sub agents. So this is actually an update from later in the video.
04:03We'll just go over this really quickly. They now have this nice sub agent tree that shows you what each and every one of your spawned sub agents is doing, all the commands they're running. As you can see, I have five agents doing 28 different tool calls right now to do this research.
04:19And then what you can see here is they're all working, and I can say, hey. Are you there? And me and Hermes can still be chatting while all those sub agents are working.
04:28So if I wanna go and say, hey. Please add on NVIDIA to that list because maybe I forgot that after, I can do that.
04:37Now that I'll spin up another sub agent to do that before I would have to wait for this entire task to finish, which might take an hour or so, now we can keep chatting, keep working while those sub agents work. So that is background agents. When you start giving your Hermes agent more complex tasks, it will, by default now, use background agents so you can keep chatting with your agent.
04:58And the other update I showed you here is a new sub agents tree where you can go in, see all your sub agents working, see all their tool calls, what they're focused on so you can keep track of things. Those are two that's a two for one update for you right there. Third one, and this is a fun one.
05:13This is not the most major of all the updates, but I have had a tremendous amount of fun the last day using this update, and that is the new Unreal Engine 5.8 MCP.
05:23For those that missed it, Unreal Engine 5.8 just released, and for the first time, they support MCP. Those who aren't familiar, Unreal Engine is, like, the most popular video game engine on planet Earth.
05:35Now you can actually use AI to build an Unreal Engine. It is a very complex big game engine, but now you can use your Hermes agent to actually build your games.
05:46While a lot of people build games using Three. Js, which is fine.
05:50That's like kind of an unserious browser engine.
05:53You can now use Unreal Engine with Hermes, which allows you to build, like, really nice three d complex first person shooters, third person shooters, whatever you wanna do, you now can build really amazing games with AI for completely free.
06:07This doesn't cost anything to do. So Unreal Engine 5.8 MCP, you now can build these amazing three d games through Hermes agent, which is sick. So next time you think about using three GS, maybe you wanna use Unreal Engine, then you can put this on the Steam store, make a million dollars, whatever you wanna do.
06:22Next up, we have an updated desktop app.
06:26If you watched one of my last videos, you know I've been using the desktop app out of the wazoo. There's so many nice quality of life improvements to this desktop app. Well, they added a ton to it in this update.
06:38First thing is is you can now open up your chats in their own windows. This is amazing for multitasking. You right click any of your sessions.
06:46You go new window. So all you need to do is right click on any of your chats, click new window, and what's gonna happen is you can open up your chats in their own separate windows. This is amazing for multitasking.
06:59You can have each window up next to each other. Have your Hermes agents go doing separate things. Watch them all at once next to each other.
07:07This is the first time you can have multiple Hermes agents up like this. Really nice quality of life improvement. Next, they added an excellent model selector at the bottom.
07:17So instead of having to do slash model or go into your config files, you can now very quickly click here and change which model you're using, which thinking level, edit any of the models. Very easy to switch models very quickly. This will save you a ton of time when new models come out.
07:32We talked to the live sub agents pane. We showed you that earlier. And then here's a nice new one.
07:36If you click this button down here, you get a built in terminal right inside of Hermes. So instead of having to open up a new terminal run commands, you can do it right inside Hermes. There's no reason to leave your Hermes window at all.
07:48Some big nice improvements to the desktop app experience. If I'm at my computer, this is the best way to be using Hermes.
07:56Next up, we are going to go to the dashboard for the new profile builder. So if you, in your terminal, run Hermes dashboard, it will open up what you see here, which is kinda your web browser experience for Hermes.
08:09We've talked about a bunch of things here before. If you watched my previous videos, you know about the Kanban board. But one thing I wanna talk about that's new with this recent update is profiles.
08:18For those who don't know, profiles are your way of spinning up multiple Hermes agents. This is something I do. I have multiple profiles.
08:26So I have my default Hermes. I also have gpt. Mez, which is my codex Hermes agent, which does a lot of my coding for me.
08:32I have librarian, which manages my memory. You can quickly spin up new profiles from here. So if we do build, we can give it a name.
08:41Say, I wanna call it coder. And we now have this really nice profile building experience that makes it really easy to build these profiles.
08:50So we go in, can choose our model. What we would had to have done before to build profiles, a whole bunch of manual work, changing config files, changing text files. Now we go through this nice onboarding experience where we can add exactly what models we want and exactly what skills we want really, really easily.
09:07So if you're not doing a multi Hermes agent approach yet, go into your Hermes dashboard, spin up new profiles, give them different models. Each model has different strengths.
09:17If you want a video on, like, which models are best for which Hermes agents, let me know down below. I'll make that next if you want it. Also, while I'm showing you this, if you learned anything so far, make sure to leave a like down below.
09:28Subscribe. Turn on notifications. All I do is make amazing videos about AI.
09:32And join the Vibe Coding Academy, number one AI community on planet Earth. Link for that down below. So new profile builder, really great experience for spinning up new Hermes agents quickly.
09:42But that takes us to the next big update here, which is a brand new skills hub, which also lives inside of our Hermes dashboard. So if we go into skills, this isn't new, the ability to see which skills you already have installed.
09:56But what's new here is the browse hub. Skills hub questionable name. But if you go in here, you can see an entire assortment of other skills you can install.
10:05So these are the featured ones, but you can also browse hundreds of other skills. So really easy to find great skills. One other thing you can do here, though, is you can click details, and you can actually do your own security scan to make sure it's safe.
10:19Now I'll tell you this. I'm not the biggest skills downloader in the world. In fact, what I typically do is I go in, I take the skill dot m d file.
10:28I give it to my agent. I say, hey. Make your own version of this and make sure it's safe for us.
10:33And what your agent will do is build your own custom version of that skill that's built for your own specific use cases. But this is a good place to go to find inspiration for other skills. Highly recommend checking out the skills hub here.
10:47Again, let's change the name of it. Alright. Let's go to back here in number seven of eight.
10:52We got two more. Smarter memory edits. This is a great one.
10:56Hermes agent now creates and updates its own skills significantly more and better than it did before.
11:04This is what makes Hermes agent so great is the fact that it's constantly self improving. It's constantly building upon itself. Now as you use Hermes in this new update, I'm sure you'll notice it actually is constantly creating and looking for skills and updating those skills.
11:21I'm actually finding way more often that it's saying self improvement review patch this skill. It's doing tons of these new self improvement reviews. You can see here, this is my Unreal Engine MCP skill.
11:33It's just constantly patching, updating, changing its skills as it does anything.
11:39I have found this has improved performance a ton. I was struggling a little bit with the Unreal Engine MCP to begin with, but after talking to it a few times, it was constantly improving its skills, and now it's excellent with Unreal Engine MCP. So there's no action items for you to do here.
11:55Just give your Hermes more prompts, more commands to do, and you will find that it updates and builds its skills a ton more. In number eight, Telegram just launched a whole bunch of new features for talking with agents, and Hermes implemented them all. There's so many new formatting features here.
12:11Check this out. So many new things your Hermes agent can do inside Telegram. It can create tables, lists, bold formatting, streaming words different ways.
12:20First thing you see was the word streaming a lot smoother. You'll see it again right there. Very, very nice.
12:26And in a second, you will see all the nice beautiful format it can do in here as well. And boom. Look at this.
12:31Look at this table it built. So it built an entire table of stocks that invested for me, gave description, stock price, market cap. So now it can do really nice things, bolding, tons of new formatting options.
12:44So don't be afraid to ask for information and data inside Telegram from your Hermes agent so you can get really nice formatted data like this. These were the biggest updates. If there's anything here you found confusing that you don't know how to implement, I got good news for you.
12:59You have super intelligence at your fingertips. You just go in, Unreal Engine 5.8 MCP. Go to your agent and say, hey.
13:06Look at the latest Hermes update. Install Unreal Engine 5.8 MCP so I can give you commands on how to build games. Boom.
13:13It'll set it up. You don't know how to open up the Hermes dashboard to find the skills hub or the profile builder? Say, Hermes, open up the skills hub.
13:20Boom. It'll do it for you. Let me know down below what kind of video around Hermes you want me to make next.
13:26I do this all based on viewer feedback. I am so appreciative that you guys would watch this video and learn from me. I just love this stuff so much.
13:34Thank you for your support. Thank you for everything you do for me, and I'll see you in the next
The Hook
The bait, then the rug-pull.
Eight updates in one release is unusual. Hermes Agent 0.17 earns the headline because the changes are structural: background delegation is now automatic, iMessage becomes a native control surface, and the agent patches its own skills without being asked.
Named ideas worth stealing.
01:43model
iMessage / Telegram / Desktop decision matrix
- iMessage = quick queries on iPhone
- Desktop app = deep work at computer
- Telegram = complex threaded sessions on the go
Three-channel model for routing AI agent interactions based on context and cognitive mode.
Steal forAny multi-channel AI agent setup where you need to decide which surface to use when
How they asked for the click.
VERBAL ASK
09:12next-video
“If you want a video on which models are best for which Hermes agents, let me know down below.”
Soft mid-video ask during profile builder demo; subscribe + like + Vibe Coding Academy hard ask in same paragraph.
MENTIONED ON CAMERA
FROM THE DESCRIPTION
PRIMARY CTAWhere the creator wants you to go next.
OTHER LINKSAlso linked in the description.
Visual structure at a glance.
