Why Hermes Agent is Outperforming OpenClaw

0:00Forbes just published an article yesterday that Hermes agent has overtaken Open Claw and I know why and in this video we're going to be breaking down why you should be using Hermes especially if you're just starting out. So you can see Hermes agentic AI overtakes Open Claw and here's the 10 shifts that leaders need to know. So Hermes agent right here has crossed over 140,000 GitHub stars which is incredible and Open Router usage data already ranks it as the most used framework in its category. So if you go to openrouter.ai/apps you could see here the top coding agents right now is a Hermes agent by 568 billion tokens being spent on this agent

0:45where if you compare that to Open Claw you're at 151 billion tokens. You got top pro productivity rankings Hermes agent and that is pretty incredible on where we have come and I know the reasons why Hermes agent is gaining so much traction. Plain and simple, it is so easy to set up and that is the hardest part when it comes to Open Claw. I've done over 40 one-on-one calls helping people install Open Claw and they've been great though it is very tedious for non-technical people to get it installed and that doesn't mean that it's installed and working because with every update it tends to break. But with Hermes agent this is an agent that is so easy to get up and going and getting your first AI automation started you can

1:32literally get this set up within 10 minutes. I have it automating my social media and as well as my email outreach for an app called The Magic Hand which is basically an AI photo editor for real estate agents. It does cold email outreach and also makes social media content that posts automatically. Also the cool thing with Hermes agent is that it has a self-recursive learning loop, which is very valuable for skills that learn and get better over time. And with Open Claw, you actually need to set this up, and it can take hours to days to actually get it to actually learn and improve its skills over time. So, Hermes agent is taking a different path. You can see its core idea is continuity, holding context

2:19across sessions, learning from prior work, and converting repeated actions into reusable capabilities. Where Open Claw that's enabled execution, Hermes is built around accumulation. And that single design choice is driving a faster shift than most of these analysis expected. So, Hermes agent launched about a month after Open Claw and has been gaining more and more traction lately. And like I said, it's basically easier to set up. So, the first thing that this Forbes article goes over is agentic timelines have basically collapsed. So, enterprise planning cycles typically assume a 12-to-24-month window for meaningful change, and Hermes compressed that into a single quarter.

3:05The result is a mismatch between planning and reality. So, if your roadmap still assumes stability, it's already outdated, and the gap will widen with every cycle you let pass. So, if you're not using Hermes agent, I really don't know what you're doing. Like, if you're in this AI space, you really need to be using a little bit of both. I still use Open Claw, but I've been using Hermes agent a lot more. You can connect it to your Codex. You can connect it to your OpenAI account. Even the $20 a month on your OpenAI account will get you so much usage. Real quick, before we keep going, if you're watching this and you want to actually build with some of these tools, not just watch videos about them, then you're going to want to check out our community down below, Shipping School. We have a full Claude code

3:50course, a full Open Cloud Course, and four live boot camps every single week where we actually help you get set up from scratch. Like actually set this thing up, not just watch a tutorial and figure it out by yourself. And we also provide one-on-one coaching, so you could book a call with me, we could share screens, and I can help you get Cloud Code or Open Claw running on your machine. That's it, no fluff. I built this community because watching YouTube's only gets you so far. We launched it just 3 days ago and we have over 55 members. You need people around you who are actually building, people who hold you accountable, and coaches who can help you when you get stuck. I'll put the link in the description down below. Get in now before the price goes up. And it's unreal, like the

4:35amount of potential that you unlock using GPT-5.5 with a Hermes agent. Like just as simple as like managing your email inbox or doing cold outreach in your voice profile. This is something that you really need to get on if you're in this AI space. Another problem with Open Claw seems to be memory. So, with Hermes agent, it uses what is called a Honcho memory. So, the agentic memory is now expected here and it says earlier systems demanded repetition. Teams had to restate workflows, preferences, structures every time they sat down to work. I'm sure you've worked on Open Claw before where it has forgot things or has lost context from your Telegram chat to your main session control UI panel. And it's been annoying and this

5:21is very frustrating. But with Hermes agent, since it builds this continuity by design, it captures patterns and decisions as they happen and then applies those patterns to future work. The value now increases the more you use it, which can change how performance gets measured in the first place. So, that is why people love Hermes is it just has better memory, straight up. The third thing they noted here is agentic learning compounds over time. So, most systems finish a task and just stop right there. But, Hermes can review outcomes, it can extract patterns, and then stores them for reuse later down the road. So, each task quietly improves the next one. And what this builds inside the system is just momentum. So,

6:07the kind that actually shows up in the output later rather than all at once. That's amazing. The fourth thing they outline here is agentic skills are becoming actual assets. So, when Hermes solves a problem, it converts the approach into a structured skill. Something that is searchable, shareable, and actually reusable. That introduces a new category of assets, one that sits apart from the code and apart from data. It is a reput- repeatable execution and the teams that recognize this early will end up with libraries of capability instead of just libraries of workflows. So, with open claw, you actually need to set this up manually, get your skills installed, make sure that there's some kind of recursive feedback loop so these

6:52skills get better with over time. But, Hermes does this automatically, which is amazing. They also say that agentic stack is becoming more clear, which is I mean true. I mean, you have a better concept of what you're going to be using this AI agent for as people start building out more workflows and generating more skills for people to use. So, this past year was very confusing, right? About what category actually requires AI agents, but the pattern is emerging here. And the stack needs to have three items in your workflow. So, identify to define who or what is acting, access to enable connection to systems and resources, and skills to determine how actually work gets done. The organizations gaining traction are aligning around these

7:38three-layer structures, and clarity is starting to compound. We're actually seeing AI agents being sold as a service, and this is the new future from SaaS going into more of AI agent architecture and just the web being built out for AI agents to use, right? So, I love this because it gives builders like us more clarity on what to build upon. I'm not saying to stop everything and give up on Open Claw. I still run my entire content pipeline with Open Claw. I've spent 2 months building it out, and I'm yet to tear it down. It still serves value. It still runs automated in the background for me. It does X research. It does podcast outreach. It creates my blog, my newsletter, and it manages my school community. So, why would I break it

8:23down? But, I think people that are building today need to use a little bit of both. It's just up to you on what you want to build. So, I've been using Hermes more more lately, and I've been having it do complete email outreach to real estate agents and tracking them in our own CRM that we built to be able to get people to sign up to our app. And that is just a easy, simple workflow that you could build out today. Whether it's managing social media, doing cold outreach, managing clients, this is something that you really need to get involved in. Like, if you look at the actual Hermes readme file, it calls Hermes the self-improving AI agent for a reason. And it says it has that built-in learning loop, which basically creates those skill files from experience. And

9:08this can take a day or two to build out in Open Claw. It improves during each use, which is the most compelling reason why to use Hermes in general. It just gets better the more that you use it. It learns more of your contents your content and your cadence. It also searches past conversations. So, it builds a model of basically who you are across each session that you use it. I use the Hermes workspace. You could download it on GitHub and get it installed. Hermes comes with its own workspace. I just like this other one a little bit better. The UI is a little cleaner, but you could also just use Telegram or whatever you want. But you really need to check out the readme file on what it specifically goes in depth on how Hermes works and what it does. Like the main most compelling feature is that

9:55self-improving learning skill. That's not a tiny feature. That's basically the whole game to this. Like we want to use AI agents that get better, right? It's like hiring an employee. What if you hired an employee that only did the bare minimum of what you're asking it to do? You want the employee to get better, right? And that's the whole point. You know, most people do not quit AI agents because the model is dumb. We're way past that now. Most of these models are really good and they're very capable. But most people quit because the agent has no memory or no rhythm of who they are. Basically no workflow and no clue what happened last week, let alone yesterday. So like you could spend an hour teaching it to work, right? And then start up a new session tomorrow and it forgets everything. And you start over. How annoying. That's

10:40exhausting and that's where most OpenClaw users have basically given up. Every time you update something breaks. It's annoying. I'm still on OpenClaw 4.21, by the way, which is like a month ago, right? So the readme inside of Hermes can run on a $5 VPS, a GPU cluster, or just a serverless infrastructure that costs nearly nothing when idle. It also says Hermes can talk through Telegram, or Discord, or Slack, WhatsApp. What Whatever kind of communication you use basically with OpenClaw, you can use with Hermes. Or you can just use the CLI as like the single gateway process if you're into that and want to just be in the terminal. It also says Hermes supports Open Router with 200 plus models, which is amazing. So News Portal, Novita AI,

11:27Nvidia, ZAI, GLM. You can use Kim and K2 or MiniMax. You can use Hugging Face as well, but I would just stick with Open AI because you get so much more token usage just on a $20 a month plan. Or you could just use your own AI endpoint if you'd like. So, if you're a builder, like the pitch is simple. Come on now. You really need to start using Hermes. Run it wherever you want, talk to it whenever you want, and use whatever model makes sense. But like I said, I highly recommend GPT 5.5 to get this going. And then, the more that you use it, the more it's going to learn from your work. And the more that you communicate it with it and talk to it about what you're trying to build, it's just going to learn and know from your cadence, your content, and build what you're trying to build and actually

12:13remember it. So, if you compare that to Open Claw, Open Claw is still huge, right? 361,000 stars on GitHub. But the the big problem is getting it set up and getting it functional. A lot of non-technical people don't want to go through the headache. Open Claw is still a monster here, and the repo is bigger. The plat platform surface is massive, and the latest Open Claw release is exactly why it became the default for most operators. If you're struggling to keep up with content, what I'm about to save you about 40 days' worth of work. I built something called Content Machine. It's 10 AI agents that run on the Open Claw orchestration, and they handle everything. Scripts, thumbnails, X posts, blogs, outreach, clips, newsletters, all of it. So, I went from

12:591,000 subscribers to 4,000 subscribers on YouTube in 7 days using this exact system. Every single morning, I wake up, and the content's already done. I spend maybe 15, 20 minutes reviewing and approving them, and I move on with my day. It works for any niche, fitness, finance, real estate, marketing, whatever you are building, and it is 100% completely customizable to your use case. So, you get the mission control dashboard, all of the cron jobs, everything I've built over the last 40 days helping me gain more and more people to subscribe and join the community. So, you plug in your own thing and it molds it to you. It learns how you talk and it writes so it doesn't sound like AI slop. $97 one time. It's

13:44not a subscription. I'll put the link down below and you'll thank me later. So, it just released 5.25 beta and in that beta they shipped iMessage attachment path fixes and duplicate iMessage watcher startup fixes. They also fixed the codex sandbox sandbox path, which is amazing, and Alpine Linux installer fixes. So, they did a lot of overhaul in these last few updates and it's worth it, right? It's not flashy but it's more infrastructure implementation and that is real operator work but still like the barrier to entry versus Open Claw versus Hermes is still a lot harder. You know, it is truly the stuff that keeps agents from breaking when actually a normal person like you and I uses this type of architecture,

14:29but the truth when you look at it is Open Claw's not dead. I think it's just built for a different type of user. The truth basically is Open Claw's becoming infrastructure and Hermes is becoming the continuity agent. That's how I see the two. And that is basically the comparison in my mind. Open Claw is the big platform that wires the models into channels and plugins and tools and workflows. It just takes a little bit more of nuance to get it all set up. But, Hermes is basically pushing hard on the memory, the learning, and that self-recursive skill feedback loop. It's always on and it's a smaller emotional promise. You know, this AI agent grows with you and I love that. And that phrase really matters. It sticks with me. My Open Claw also grows with me but

15:16I had to build that out. Like I have it going to my YouTube channel via the API and building actual feedback loop of what content's performing. Same thing on X, too. It reads the tweets and sees which one performs with the right hooks, but I had to build that out as well. It's not easy. You know, people don't want another tool, you know, they want just Hermes to work. They want Open Claude to work. And that's the part you've seen with these updates, especially with Open Claude, they're just trying to build the infrastructure so it's more reliable. So, if you go to the actual Hermes page, they say that they have 237 stories across 15 categories. And if you look here, it's pretty impressive. I'll put the link to this down below. Like you really need to check out what people are building. It

16:01is incredible. User stories and use cases, what the Hermes agent community is actually building. So, every tile below links to an actual real post or an issue or a video of basically what people are building with Hermes. I mean, it is pretty cool. I I just love this because it it shows real-world application of what developers and non-technical people are building with these agents. And like you can see here, someone posted told it to Google me and ship a landing page to my VPS. I mean, that's amazing. That's awesome. And you see here a skill audit skill that improves itself on a cron job. You need to remember, too, when we talk about AI agents, don't think it's some mystery. All AI agents are are cron job in a markdown file. That's it. So, really

16:47cool. Like I said, I'll bookmark this down below. And it's just really cool to see. Also, the Open Claude community and, you know, Hermes agent building amazing things. If you go to our Shipping School website, there's 103 Open Claude use cases here, what people are actually building. And like I said, I go back to just being able to see other people that are non-technical build with these agents should give you motivation and inspiration to build something yourself because there's no magic of just getting your hands dirty and start playing around with this technology. So, like I said, user stories, what the Hermes agent community is actually building, this is a beautiful thing. So, like there's so many stories out there, right? And another story that I had come across on the Hermes page is a user turned $100

17:33into $216 in less than 48 hours with basically a self-learning weather trading bot that scanned weather markets basically every 16 minutes. And then it took that info, it compared three forecast sources per location, and then it bought undervalued temperature buckets. I didn't even know you could do this. And then it flipped those for a profit. It reviewed what worked, it wrote strategy notes, and then it adjusted them for the next time they made the investment, which is pretty impressive. Another story on that site says a user used Codex with GPT 5.5 on extra high to watch Hermes agent to agent workflows run, and then catch

18:19where that workflow broke, and then fix it live until it worked reliably. I mean, that's an agent workflow that we would all love to have. And the last one that I came across, and the pattern that I actually care about the most, because the future isn't not one agent doing everything, it's multiple, and the future is basically agents supervising other agents. Like having an orchestrator agent and having agents underneath it, the agent finishes the work. I don't know if I'm portraying this correctly, but the agent's going to report back to the orchestrator agent. The orchestrator agent's going to check the work of that agent to make sure it actually completed, it did the work, and verified it. That's how it works. We need a supervisor agent in the loop before the human in the loop comes and reviews everything, right? So, basically

19:04it's an agent watching another agent making sure it's finishing. So, another agent can like check the logs, another agent can turn the fix into a skill, and then another agent can schedule the next run, and that's amazing. That's literally how builders are wiring these systems together now, and according to the Hermes user page, that's how they're working. You know, you really need to check it out. It's updated every day, by the way. And here's what I think builders should be doing with this information. Do not pick a team. This isn't sports. This isn't like, oh, I'm married to Open Claw, or I'm married to Hermes, or I'm married to Nemo Claw. there's so many out there. Use them all. Find what's going to work for you. We're all building different businesses here. You know, that's amateur hour. Like, basically thinking like, oh, this is like, I'm married to this one

19:49technology. I'm married to Codex. I'm married to Cloud Code. I think we need to be using all of the tools just to make sure what's going to be the best for our business. Like, use Open Claw. Get it going. Use Open Claw when the job needs a lots of channels, or you need broad platform surface, or you need plugins that go deep. You need mature operational plumbing, or you need the kind of boring, reliably fixes that actually show up in release notes in 5.25, or study Hermes. You know, when the job depends on basically long-running memory, or skill creation, or you want the remote, always-on agents, or model flexibility, or you just want workflows that get better over time, play around with Hermes. And if

20:34you are building for clients and trying to sell agents as a service, this is where it gets a little more interesting. You know, clients don't really care about whether using Open Claw, Nemo Claw, or Hermes agent, or Cloud Code, or Codex. They don't care about that. They just care whether the invoice got reconciled, and that your actual workflow creates the desired outcome. You know, basically, like, if you're doing something with lead follow-up, they want to make sure that the lead actually got followed up with. And like, if you're generating reports, they want to make sure that the report actually got sent, or the bug got fixed, or the video got drafted. It is just simple things like that. They want to make sure it works. So, your job as a builder is not to memorize the logos and be a fanboy about everything that comes out. Your job is just to basically build repeatable workflows that work. You need

21:19identity, who is the agent? You need access. Well, use one of these tools. Like what tools and accounts can it touch? You need to figure out what is the best tool for the job. You need skills? Well, start developing skills. And for Hermes agent, they does it automatically. I really think people are just stuck on prompts, you know? They're just like trying to figure out what is the right prompt to send. They think the game is just prompt harder. We're going to be way past that. I do think that prompts are important, but they're not end-all be-all. The game is to give the agent you're working with a job and then give it the right tools, give it the right memory, and give it constraints, and then watch what it does. And then turn the repeatable workflow into a skill, and that's where all this is

22:04going. The Hermes readme says it has scheduled automations with built-in cron and delivery to any platform, daily reports, nightly backups, and weekly audits running unattended. I love that. It can also spawn isolated sub-agents for parallel workflows. So, that's amazing. It has seven terminal backends, so you could do local, Docker, you could do SSH, Singularity, Modal, Daytona, and Vercel Sandbox, which is pretty cool. And that's a serious agent with serious architecture. But Open Claw is not sitting still because they are still shipping tons of features. In the newest update, they fixed MCP tools, and they do a lot of back-end infrastructure

22:49that's not fancy, you know, it's not going to make an amazing video, but it is becoming more reliable. Going back to the Forbes article, agentic control is becoming the differentiator, and I totally agree. Hermes is designed to run inside infrastructure that companies already control, and that's why people are choosing it because it's just more simple to get going. You know, it's using their own models and their own data, and that gives organizations ownership of how work is executed all the way down. So, I love that. People are just plug and play, ready to go. Agentic infrastructure decisions are back. So, Nvidia and others are positioning their systems as the foundation for continuous execution by governments and requesting and facing leaders is no longer just which system

23:34to use. It is where the system should run. I love it. I mean, you see big players in the game, Nvidia obviously released Nemo Claw, their own version. This is where things are headed. They have number eight, agentic oversight is evolving. So, running systems locally reduces certain risks and introduces new ones. When a system is learning, storing patterns, and acting across tools, visibility over time becomes a own discipline. So, it's basically saying oversight is shifting from initial access control towards something more continuous and closer to ongoing accountability than just a one-time approval. That's why I think agents managing managing agents is the next big boom in this cycle. Agentic deployments generate leverage. These systems do more

24:21than finish tasks. They produce structured records of how work actually got done. Those records become raw material for refining processes, improving performance, and testing new approaches. Think of it like an SOP, right? You have an SOP in your business, give it to an agent, and that SOP can improve over time with refinement. The 10th and final one is agentic speed is now the strategy. So, Open Claw helped define the early market just about 4 months ago, but Hermes sped that pace up. It improved faster, adapted faster, and spread much faster than Open Claw, and then the market rewarded all of it. I think the competition between Open Claw and Hermes makes each agent better. Speed itself has become the strategy and that reframes how leadership teams

25:06should be thinking about adoption windows. So, if you look at all of what's going on in the last few months and you're not using AI agents, if it's Hermes or Open Claw or using Codex at personal computer or using Claude code on schedule cron jobs, I really don't know what you're doing even watching this video. You need to start playing around with these tools and if you need help getting these tools working, we do a four live boot camp call inside of our community where we teach courses and deep tutorials from deep technical knowledge to just basic what is an API call, what is an MCP, getting your first AI agent installed and running. Check out the link down below. Like I said, we do four calls a week. We also have one-on-one calls that if you wanted to schedule, we're happy to do that as

25:52well. So, check out the call check out the community. We'd love to have you and what are your thoughts? Where do you see this AI agent race going? And we'll see you in the next video. Have a blessed day. Bye-bye.