← Back to search

Did Hermes Agent just kill OpenClaw? (full guide)

Alex Finn · 2026-10-30 · 13 min
relevance 100 2757 words Episode page ↗ Audio ↗
Show full episode description
Hermes Agent is the newest open source competitor to OpenClaw. It is self improving and REALLY fast. But is it better?FULL Hermes bootcamp in the Vibe Coding Academy coming up: https://vibecodingacademy.devSign up for my free newsletter: https://www.alexfinn.ai/subscribeFollow my X: https://x.com/AlexFinnMy $300k/yr AI app: https://www.creatorbuddy.io/Hermes Agent:hermes-agent.nousresearch.comTimestamps:0:00 Intro0:30 What is Hermes Agent1:37 Differences with OpenClaw2:31 Hermes demonstration6:10 What OpenClaw does better9:22 Best workflow Learn more about your ad choices. Visit megaphone.fm/adchoices
✨ Episode Outline — click any point to jump to it in the episode
Problem solved
Whether Hermes Agent should replace OpenClaw, and how to set up and combine both for maximum productivity.
Benefits
  • Hermes is noticeably faster and more lightweight than OpenClaw on the same model (Opus 4.6)
  • Self-improving: Hermes turns completed tasks into reusable skills automatically
  • Built for tinkerers with ML/RL tools and first-class open-model support (Qwen 3.5)
  • Two-agent setup acts as an insurance policy: one agent fixes the other when it breaks
  • Full transparency into every tool call and step as the agent works
Use cases
  • Daily 9am cron: Hermes pulls top 5 Hacker News stories, scores them for AI/startup relevance, generates an audio briefing, delivers to Telegram, then creates a 'Hacker News Daily AI Briefing' skill
  • OpenClaw as main orchestrator spinning up Hermes via ACP for tasks it's better suited to
  • Side-by-side multitasking: Telegram (OpenClaw) and Hermes CLI working on different parts of a codebase
  • App build: Opus-powered OpenClaw on front end, ChatGPT-powered Hermes on back end
  • Obsidian-based memory system fixed Hermes memory issues and improved OpenClaw memory too
KPIs / results
  • Used hardcore for several weeks before verdict
  • Vibe Coding Academy: 1400+ members, live Hermes bootcamp
  • Recommended 15-20 minutes/day tinkering with new tools
Tools / build
0:00 / 0:00
Hermes Agent is taking the internet by storm. A lot of people are even claiming it's better than OpenClaw. Well, I've been using it hardcore for the last few weeks. In this video, I'm going to cover what Hermes Agent is, what makes it different than OpenClaw, how to set it up and give you some incredible workflows with Hermes Agent that will absolutely explode your productivity. I will then let you know if you should be switching from OpenClaw. Let's get into it. So here it is, Hermes Agent is running in my CLI. You can also run it in Telegram and Discord and WhatsApp and all the other messaging apps, just like OpenClaw. It is a lot like OpenClaw in many different ways, but it's also very different than OpenClaw in some very key ways. And if you're brand new to OpenClaw or my channel, Hermes is an autonomous AI agent that just like OpenClaw can control your entire computer. Anything you tell it to do, it can do it on your computer for you. Automate incredible workflows, all from the platforms you're used to like iMessage, WhatsApp, Telegram, or Discord. It's like your own personal AI employee. So what I'll go through first is from a high level, what the difference is between Hermes and OpenClaw. Then I'll go through a full demo and tell you how to be using each. By the way, Hermes is very aesthetically pleasing. It does a lot of really cool visual tricks. It even can create sick ASCII art video automatically. You just say, hey, can you make me an ASCII art video? And this is what it came up with. I know this is a little bit trick, but it does some amazing things. Stick with me here. So here's actually a chart it built. So it has built in skills, including Excalibra skills. So it actually made this chart for me. I'd edit a couple of things, but it came out really nice here. Hermes versus OpenClaw. What makes Hermes agent so good? Well, a couple of things. It is incredibly lightweight and performant. It moves so much faster than OpenClaw. Even on the same model, I have Opus 4-6 on both. Hermes moves much faster. That's the first thing I noticed right off the bat. I've had a couple performance issues with OpenClaw over the last couple of weeks. I've been able to sort them all out and I'll have a video about performance and OpenClaw coming out soon. But Hermes over the last few weeks has been super quick and speedy and very, very performant. Here might be the biggest thing everyone's using it for that it's definitely the most impressive. It is very, very self-improving. They put a huge focus. So let me give you an example here. So let's go into Hermes. I'll give you a nice prompt to demonstrate this. I'm going to do is give a pretty complex task here. I'm going to tell Hermes every morning at 9am, I want you to pull a top five hacker news stories, fetch the actual articles, summarize each one, score them with relevance to AI and startups, since that's kind of my niche, generate an audio briefing of the top three ranked ones, and deliver it to me on Telegram with a text summary alongside the audio. So basically, I'm getting a briefing every morning that gives me a custom podcast and brief on all the top AI stories. Let's hit enter on this and I will show you the self-improvement in action, how it figured things out and how it builds its own skills on the fly. So it can do anything it wants. So this is sick. You can see it do this task step by step. One thing I love about Hermes is it is completely transparent. You can see every little step it does, every tool call it does. You can see it going to each specific website, taking snapshots, commentating on exactly what it's doing. You can see it finding the stories as it goes. I love this. It allows you to see exactly how Hermes thinks and how it works. Now look at this. This is where the self-improvement comes in. Now it is capturing everything it learned and turning it into a reusable procedure. So you can see here, Hacker News Daily AI Briefing Skill Created. Now it's updating the cron job to reference this new skill. And it says, boom, let me deliver the actual briefing to you on Telegram. Says it sent it. Let's check Telegram here. I'll pull Telegram over and boom, look at that. There's the briefing. There's the five stories. Let's hit play here. Here's your Hacker News Briefing for Monday, March 30th, 2026. I hope you can hear this. I hope the audio is going through. Story number one and the most relevant to you today, UniversalClaw.md, a new open source project that claims to cut claw. It's like my own personal morning podcast. I basically have a personal podcast every morning. It's giving me all the news I need. So as I'm off to work, I can press play on this and hear personal podcast. That's pretty amazing. But you can see here how the self-improvement works. It plays around, it tries different things, it sees what works, and then it cements everything into its own new skill set. So the more you use it, the better it gets. OpenClaw is fantastic at figuring things out as it goes along, but it really seems like Hermes has put an emphasis on the creating of its own skills and the self-improvement. Let's go back here. We talked about how it's lightweight and performant. We talked about self-improvement. Couple other things that are really cool about Hermes before we talk about the differences with Open Claw. It is built for tinkerers. So it has a lot of really cool technical tools built in. It has machine learning tools built in. It has reinforcement learning tools built in. That might sound super complex, but basically it's built for AI tinkerers, people who want to fine-tune their own models and do things like that. It has built-in tools for that, which is really, really cool. And it's also built for open models. A lot of people are running their Hermes agents a hundred percent on open models, a hundred percent on Quint 3-5. So if you listen to my channel a lot, if you watch my live streams, you know I'm huge into open models. I'm huge into people buying hardware to run their open models. Hermes has said on record they support open models. They want to build tons of tools for it. Open Claw on record has said they do not recommend open models. So if you're more of a tinkerer, if you're someone who wants to experiment with AI models and play around with them, Hermes is building a ton of tools for that. Now on the other side, Open Claw still has a lot of strengths. And again, in a second, I'm going to go through the workflow of how to use these together, what the ideal workflow is, how to explode your productivity with this. If you learned anything so far, make sure to leave a like down below, subscribe and turn on notifications. All I do is make amazing videos about AI. Also, I'm doing a live Hermes bootcamp next week in the Vibe Coding Academy. So if you're interested in Hermes, hit the link down below for the Vibe Coding Academy. It's the number one AI community on planet earth. There's over 1400 people in it. Live bootcamp next week. Also, if you join right now, you get like the last eight bootcamp recordings, but Open Claw still has a lot of strengths. There's still a lot of great things about Open Claw. There is an absolutely massive team working on Open Claw. Obviously, Open AI, Aqua hired the creator of Open Claw, Peter Steinberg and are providing resources for Open Claw. On top of that, this is not talked about as much. Nvidia is providing a tremendous amount of development resources for Open Claw too. So they are shipping daily updates a ton. It is constantly transforming. That is an advantage to OpenClaw. They're able to ship tons of updates a lot more quick. It is a little bit more stable. Hermes Agents, I had a couple of issues I was able to work through. I had a memory issue that I fixed by implementing an Obsidian system. And actually this Obsidian system improved both the Hermes memory and the Open Claw memory. If you're interested in using Obsidian with OpenClaw and Hermes, let me know down below. I'll probably make that my next video if there's interest around that. But it seems like Open Claw might be more the Hermes issues I was able to work through and it works fine for me now. There is a larger community behind Open Claw at the moment. So they're able to get a lot more open source work behind that. Although I know Hermes is blowing up right now and is catching up. A ton of people are talking about Hermes on Twitter. And there's also some smaller things like they built out native plugin support across cursor and Claude code and everything for Open Claw, which is huge as well. Point being is there are advantages to both. They made Hermes in a very specific way that made it very cool for what it does. The self-improvement's amazing. It's constantly getting better the more you use it and it's very lightweight and performant and it's built for tinkerers. At the same time, Open Claw has tons of features that are constantly releasing. So the question becomes this, should you use Hermes Agent or should you use OpenClaw? Is Hermes Agent good enough to rip out Open Claw? Well, here's the answer. Here's the truth. You should a hundred percent be using them together. There are so many advantages to using them together. They complement each other so well. And here's exactly how I do that. I use Open Claw as my main orchestrator agent. When I have complex tasks I want my agents to do, I'll say, Hey, Open Claw, here's the task I need you to do. Feel free to spin up Hermes as well. And it will figure out what tasks OpenClaw is fit for, what tasks Hermes is fit for. And it will actually spin up Hermes Agent to do those tasks. Another incredible way to use these two is by using them side by side to multitask. So I will literally have Telegram and my Hermes CLI next to each other, side by side, just giving them different commands each, having them work on different parts of the code base, having them work on different parts of projects. It has made me so much more productive. Instead of having to wait around and watch one agent do something, I can have them both going at the same exact time. I'm never doom scrolling or distracted like most people are when they're waiting for their agent to do something. To allow Open Claw to communicate with Hermes, you just need to set up ACP. All that is, is basically a protocol that allows agents to communicate with each other. You just go into Open Claw and you say, Hey, please set up ACP with Hermes Agent. And it will go and just set up for you. It's very easy to do. And then it can start sending messages directly to Hermes. I'm building an app right now. It's a very special project I'll be announcing soon. But I have both these agents working on at the same time. While I have Henry, who's powered by Opus, working on the front end of the app because Opus is a little bit better with front end code. Then I have Hermes coding agent, which is powered by ChatGBT working on the back end. This is a very powerful workflow. One being powered by Claw, the other being powered by ChatGBT working on the parts of the app that they specialize in. And the beauty is the more you use Hermes, the better and better and better it'll get. So maybe one day it'll overtake Open Claw. I don't know. I've been using it for a few weeks. It's definitely gotten better. It's definitely still very, very, very fast and performant, but I am not ripping out Open Claw just yet. Having both work side by side has been incredible. Here's the most important thing in them all. If you are into AI and you're trying to get ahead and you're trying to create more economic value, you need to be using the latest tools. Hermes agent is that latest tool. Whether you end up using it full time or not, you have to have at least 15 to 20 minutes a day where you're just tinkering with things, trying new things out. Hermes agent is that thing that you need to be tinkering with and trying out right now. It's compatible with all the same models as Open Claw. You can connect it to ChatGBT, Open Clawed, Minimax, Kimmy, whatever you want, you can do the same things. You can try out a cheap model to see what it's like. But I think having this multi-agent workflow is really powerful. Here's another thing I forgot to mention too, which is really, really powerful. A lot of the times I will have an issue with Open Claw. I'll even have an issue with Hermes. When you're running a single agent workflow and it goes down, it becomes really difficult and kind of stressful to get it back up because you don't really have anything to talk to to fix it. You have to kind of like look at the config yourself or open up Claude code. But now that you have a two agent system, Open Claw and Hermes, if one goes down, you can go to the other and say, hey, the other one's down. Can you fix it? So you kind of have like an insurance policy as well, where if anything goes down, you have another helper there to fix it and bring it back up. So having that two agent system just gives it another layer of reliability. I've also had some issues with both of these where that maybe there's an update and it breaks a piece of functionality. Having that second agent there is great for fixing it too. It's just a much cleaner, easier, safer workflow, having those two agents available. As for which model to use, I've been saying it for a long time, Claude is the most agentic model out there. It's just the best at figuring out how to complete tasks. I highly recommend using Claude. If you want to save money, ChadGBT is still great. And if you're a little bit more advanced tinkerer like I am, you can even plug in your own local models, hook in Quen 3-5 locally. A lot of the local AI model community has been doing that using Quen with Hermes agent and they've said they've gotten great results. I need to test that as well. I hope this was helpful and I hope you learned a lot. If you learned anything, I'll leave a like down below, subscribe, turn on notifications. I'm going to have another video this week about Obsidian if you want that. I've been playing around with this really cool tool called Paperclip. I will probably have a video about that as well. Let me know which you're most interested in, how I can help you out the most. My only goal with this channel is help you out as much as I can, educate you as much as I can, help people embrace these incredible new pieces of technology that are dropping left and right. It really is the most exciting time ever to be live and be in technology and be a tinkerer. It's so much fun, so exciting. I'm so grateful for you watching my videos, my live streams if you do that. Sign up for the Vibe Code Academy down below if you want to ask me questions directly, totally up to you and I will see you in the next video.