Everclaw & The End Of SAAS | Agents That Never Sleep
59 minAudio
Zion, hear me! For Season 2 of the Mere Morpheus podcast I'm going to be uploading the Morpheus builder weekly updates. My reason for doing this is to make it more accessible for those who want an update directly via their podcast app and to have them all in one place. If you'd like to participate they are held weekly on the @MorpheusAIs Twitter spaces. So enjoy Episode #62! Recorded: 12 February 2026 AI SUMMARY: This episode features David unpacking Everclaw an open-source skill bundle to keep OpenClaw agents online "forever" with Morpheus protocol integration. Everclaw removes friction by auto-configuring Morpheus inference managing ongoing inference with a swap-and-stake flow for MOR tokens and stabilising connections with the Gateway Guardian. David outlines a clear setup path for new and existing users covering security basics whitelisted comms persona files automations and model routing for cost and capability balance. Real-world use cases include 24/7 personal agents yield-farming experiments and replacing SaaS with self-hosted workflows. The discussion covers VPS vs Mac mini trade-offs secrets management reputation rails agentic payments and why open user-owned inference matters. It concludes by exploring why OpenClaw is a lasting framework how community-built skills drive open-source improvement and why now is the time to onboard people to agents they truly own. Timeline: (00:00:00) Intro (00:00:40) Introducing Everclaw: a quick "skill" for OpenClaw (00:03:20) Packaging the fixes: Gateway Guardian and Everclaw bundle (00:07:06) Hardware choices: why many pick a Mac mini (00:10:26) Security first steps and setup guide flow (00:12:59) Personalisation: user.md, soul.md and cron jobs (00:15:11) Community Q&A and acknowledgements (00:17:28) Money agents: yield farming ambitions and wallet hygiene (00:23:43) Replacing SaaS: agents code and host for you (00:26:03) Career advice in an agentic era (00:29:14) Model choice matters and adding a routing skill (00:32:05) From posts to conversations: agents talking to agents (00:36:48) How much access? Assistant vs you, and guardrails (00:40:17) Firewalls for tone and facts; controlled autonomy (00:44:43) Swarms and pipelines: human‑in‑the‑loop for now (00:48:01) Own your inference: why this moment fits Morpheus (00:52:15) Moats and the open‑source singularity loop (00:56:14) Crossing the chasm: personal agents arrive (00:58:01) Call to action: contribute to Morpheus and setup guides If you would like to support me for the value I provide you can use the referral link below when depositing capital and we will both earn bonus MOR. Alternatively you could send me a token of appreciation directly to my wallet ;) Referral Link: https://dashboard.mor.org/capital?referrer=0xE935f231c99c04Ee0b4532a3d0BdA81B152a0384 Wallet Address: 0xE935f231c99c04Ee0b4532a3d0BdA81B152a0384 Connect with Mere Mortals: Website: https://www.meremortalspodcasts.com/ Discord: https://discord.gg/dX9BEhT3kz Twitter/X: https://twitter.com/meremortalspods Instagram: https://www.instagram.com/meremortalspodcasts/ TikTok: https://www.tiktok.com/@meremortalspodcasts Value 4 Value Support: Website: https://www.meremortalspodcasts.com/support Paypal: https://www.paypal.com/paypalme/meremortalspodcast
Audio and video play straight from the show's own hosting. PodSubs stores and serves no media.
Transcript
From the show. Tap a time to play from there.
David JohnstonBo, do you wanna kick us off, brother?
Bowen BelmerYeah. Absolutely. So, you know, the goal with this call has kind of shifted, over well, really kicking off 2026 to being less focused on building Morpheus the protocol and more on using Morpheus the protocol and building things on top of it and things that interact with it. So because of that, we've, you know, target having having various builders on as frequently as we can to talk about the things that they're building and how those things integrate with Morpheus and maybe how they can integrate with Morpheus if they don't quite yet.
This week though, David, is joining us to cover a I I don't know what to call it, David, like a mini project. I mean, it's definitely a project, but it was, you know, delivered fairly quickly, called Everclaw, which I will let you introduce, David, and and, you know, why we're all here and kind of what the thought process was behind this and what it delivers to Morpheus ecosystem.
David JohnstonSure. I think Everclaw can be best described as a skill. Right? And that's the language and the terminology that people are using related to OpenCLaw. Right? So somebody creates a new security checker, and they'll release it as a skill, and they'll post it to, Claw Hub. Right? And then all the agents can see that, oh, there's this new skill. It might be useful, and, they can integrate it. Because maybe kinda back up a little bit. A lot of us have been on this same journey
where we first set up our OpenClaw, but it had a bunch of problems. Right? Like, first of all, you needed a clawed API key, and most, you know, nonengineers didn't have a claw account. They didn't have a clawed clawed account. Right? So, you know, the first one of the first things they did was integrate the Morpheus, gateway API. And so now it just grabs the key, right, and can you can get started with OpenCLOA without having to have a clawed account. Right? So right there, that's, like, instantly removing friction,
for people. The second thing is I really loved using my OpenClaw, but, like, a couple hours in, it would just run out of credits, right, and stop responding. And so it's like, okay. Well, how do I get it to run forever? Well, wait a minute. You know, I've got Morpheus, so I can buy some more. So I created a swap skill. So you can start with USDC or or ETH or whatever, and it'll swap for Morpheus. And once you have your Morpheus tokens, it'll stake them,
and you've got inference. Right? All of a sudden, you can access inference. The other aspect that I saw was sort of a really big friction point for people was the the gateway itself. Like, your connection to your your agent kept randomly turning off and closing. Right? So I ended up developing a skill called gateway guardian. And, basically, what it does is check that your connection to your agent is online. And if it's not, it automatically restarts it. So, you know,
I I've been sort of one by one, if you will, debugging why my OpenCLaw wasn't working all of the time. And sometimes it was the gateway, and sometimes I had run out of inference, Right? Or or I didn't have access to the right accounts. And so with each of these skills, I sort of just wrapped them into a package and gave it the name EverClaw so that when somebody wants to fix this problem, like, how do I keep my my agent online forever? They have one thing they can download.
And it has some other aspects we can dive into on on security and other stuff, but I'll I'll kinda stop there. I think it would be great to sort of just ask questions about it because so many people are asking me how do I set up my my Everclaw, and I'm I'm hoping this space is something I can point them towards and be like, you know, if you're gonna do it, you know, go to everclaw.xyz, and you can download, you know, the skill, and it'll make your life a lot easier. Right? Maybe you can get your your system set up in thirty minutes instead of six hours.
Bowen BelmerYeah. Absolutely. I mean, so this the first thought that I had actually when I saw people starting to use this was the economic piece and the ability to have, like, continuous inference so that you're not accidentally making an API call that's costing you, you know, hundreds of dollars in credits perpetually by mistake. So I I thought that was, you know, a a really great idea, and I'm really glad that you did this. David, can you talk a little bit about
I I'm trying to think of like how how granular we should go to the floor of somebody setting something up like this. Because first of all, I would imagine we kind of have two camps of people right now. We have the people who are just exploring, you know, downloading OpenClaw for the first time and and setting up their initial agent and going through what that initial setup looks like. Ideally, you know, people like myself will set up the very first time straight into
EverCloud, use Morpheus right out of the gate. And then we have people who are probably more tech savvy, who maybe have an instance set up currently, whether it's on a a VPS or on their local device, who might be looking to switch over switch over to using EverClaw. I'm trying to think through the the best process of kind of tackling both of those people, maybe starting with the entry person and then working our way to the more technical route. Sure. Yeah. Well and and that's the same way I think about it. And so when I set up the EverQuote website,
David Johnstonit it serves both of those personas. So you can click on one button like, hey. I don't even have Openclaw, and it will help you download Openclaw and then give you these skills. Right? So you start with Morpheus day one, and then there's a second button near the top of the page. Hey. I already have OpenClaw. Just add this skill. Right? And so you click on that button, and it'll show you the commands. And all you have to do if you've already got OpenClaw is drag and drop them in to your chat with your agent, and it it'll figure out the rest. Right? So why don't I I agree. Why don't we start with, like, you don't have
OpenClaw yet? Right? So let's start with the the people because there's many more people that are rushing to download it this week than, you know, the early adopters last week. Right? So, if you don't already have it, I would go to, everclaw.xyz. I would click the button that says, you don't have Openclaw yet. It'll download that on your system. But even before you start that, why people are rushing out and getting Mac minis.
And I think this is important. They're doing it because they wanna set up a personal AI, but they don't necessarily wanna mix it with all of their existing, you know, files and, you know, accounts and everything. And I think that's wise. Right? OpenClaw is still super early. Right? There's probably gonna be a lot of security stuff that needs to be addressed. And so I think the smart and safe thing to do is to have a dedicated machine. And the truth is Mac Mini is not that bad. It's it's $600.
Right? Walk into an Apple store. My my friend walked into the Apple store to get his the other day, and he's like, yeah. The people at the store had no idea why all of a sudden Mac minis were so popular. Right? But they're they're gonna find out. And that's just a really good solution for sort of segregating what you're doing with, you know, your your your OpenClaw agent and sort of all the rest of your digital life. Right? So, you know, it has a little bit of cost, but I think it's worth it to have a dedicated machine.
Also, you kinda want something you could just plug in, and it's always sitting there and running. Like, if I put it on my laptop and my laptop battery runs out, it stops. Right? And so having a a machine that started just sitting in the corner plugged in, working all the time kinda goes with this cool idea that, my agent can be working for me twenty four seven, three sixty five. Right? While I'm sleeping, while on vacation, whatever, it's still, you know, going and trying to accomplish my goals. Right? So so I would say,
you know, one, it's I think it's worth getting the Mac mini. Right? If you don't have the $600, there are other options. You could put it on a you could probably put it on an old device. One of the things to understand is it doesn't take a lot of compute. I had a friend who put this on a 2014 Mac Mini and it worked. Right? The actual OpenCLO package isn't that big. It's not a big program. It's not really heavy because what you have to understand is most of the compute is being done from the inference.
And so if you give it a Morpheus API key, you know, the the the work is being done by whoever's hosting that Kimi 2.5 model or the GLM model, whatever you connect it to. Right? So you don't need the local machine. You don't you do not need unless unless you want, like, perfect local privacy, you know, to drop 10 k. You do not need to drop most people do not need to drop 10 k on a Mac Studio, right, and be able to run all of the inference. Maybe that's ideal, but most people don't have that much cast sitting around.
If you just wanna get started, the I think the Mac Mini is a great way to get started. Right? And it comes with so much of what people are used to, then you can just talk to your agent on iMessage. And it has all of the programs built in and, you know, all the things you would need to get started. It's also sort of a lot faster and easier. So I think it's think for most people, most non engineers, Mac Mini is a great easy way to get OpenClog going.
But there's still, like we talked about, a lot of security questions. So that was the big focus of last week. And so that's kind of the first set of things I put into EverClaw were these skills where it would check the prompts. It would check the skill registry that there was nothing malicious. And so I think the security's come a long way just in the last week because as everybody rushed in, they work to improve it. So I think it's a pretty good point at at this level. So if I I'm gonna turn this into a guide.
Step one, buy your Mac Mini. Either or or or just grab an old device. If you got an older one or or some old computer, you probably run it on that. Right? But go and grab a Mac Mini. Step one. Right? Step two, go to the Everclaw website. Go ahead and install, you know, the version where you don't already have, Openclaw. The next decision, step three that you're gonna have to make is how do you talk to your agent? So personally, I'm a big fan of things like Signal, which are encrypted end to end,
but iMessage isn't a bad, choice either. It's actually pretty decent compared to, like, WhatsApp or or some of the other messengers. So let's say you're using Signal or iMessage. What you wanna do is set up your agent only to talk to you. Right? A lot of the attack vectors are, like, if you just expose your agent to the entire Internet. So you're gonna set you you just tell it, hey. Whitelist only my number, only my, you know, signal ID or whatever. And, David, can you tell it that just in natural language? Can you literally just say, hey. I'm the only person you talk to? Yep.
And and as part of the setup guide is picking a communications channel. Right? So as part of the setup guide, like, you pick your inference provider. Right? But with Everclaw now, you've got that automatically set up through the API gateway. Like, great. Now I now I could at least get started. Right? And then, yes, in the startup guide, it's like, okay. What communication channel do you wanna enable? And I enabled, you know, a secure end to end option, And then, you know, it tests it itself. Right? That's the cool thing about this. You can literally just tell it what you want.
But you have to do these sort of basic setup steps like, alright. Where are you connecting for inference? Morpheus? Cool. Got it. Where where do you wanna connect, your agent, signal, iMessage? Wonderful. You know, go through, you know, all of that, setup wizard, and you'll be most of the way there. Right? And and it has other cool stuff. Like, I started it it has a dedicated, let's say, step four. You know, it has a dedicated file about you. How do you want your agent to talk to you? What are your goals? What are your values? What are your principles? What are the things it should do, and what are the things it should never do? Right? So I spent, I don't know, a good amount of time, maybe an hour, kind of explaining my core ethos.
And it saved all of that to a user dot m d file. And when it talks to me and it and interacts with me, it has that context for how I operate, what's my my theory of how to to get things done in the world, what values are important to me. It also has a sole dot m d file that defines itself. Right? So if you want to give it more context on how it should work, there's a place for that. So so, yeah, that's sort of step four and five is is
maybe step five is it's it's got all these cron jobs that you can run. So running, you know, hey. Give me a reminder about all these birthdays that are coming up. Right? And do it a week before, and let's look at presents and give me recommendations for presents for each person. Like, all that kind of stuff that runs in the back of your mind or maybe you have a calendar invite or something, it becomes way more interactive once you tell your agent. So I've gotta give them daily reports every morning.
You know? After, you know, the kids are off to school, I I grab you know, I've got a daily report about everything that's going on. Right? Or I have it, you know, scanning the news for particular things, or I've got it looking out for security, stuff that I should upgrade. Right? I even have it run that every day, a report about new skills that are available, and then I can add those skills, to make it more powerful. So I've got, like, this loop going every day where it helps me with my productivity
and it do the things I wanna do, and I spend a little time improving it. Right? So it continues to become more interesting and powerful. I think, David, that's a really good place to pause and and try to inject some other speakers. Great question. Oh, David, can you hear me? One thing I I I do wanna give credit for is, you know, Kyle did an amazing job on the API gateway. And when I went to add that as a skill, like, it was literally one, two prompts,
and fifteen to thirty minutes of, you know, the agent, you know, just crank it away and boom. Here you go. Yeah. Everything's set up. And and that was to produce the skill. If somebody, you know, downloads this and and uses the skill, it's it's nearly instantaneous. Right? You know, five, ten minutes for it to set up all its, new parameters. Hey. So, you know, the folks that have done the hard work in Morpheus the last couple years, now we get the benefit. David, can you hear me okay?
Bowen BelmerYeah. I can hear you. Oh, okay. Perfect. Sorry. I didn't mean to interject, but I just wanted to stop because I thought that the point we were just at was So, yeah, it's been it's
David Johnstonbeen very it's been very wild to see see how it evolves.
Bowen BelmerHang on one sec, guys. I think we might be having an issue with spaces where we cannot hear you. No. I can't hear you both here. Yeah. Don't think David Oh, David is on a delay.
David JohnstonI'll go ahead and rejoin.
Bowen BelmerOkay. Well, he does that. It's actually a good time because I was gonna open it up because I wanna hear from any other community members that we have in the chat as well. Is on on the topic of, like, thinking through use cases for this thing, this is what I think is one of the funnier parts of this and certainly something that I relate to personally is, you know, we've all wanted access to super intelligence and and agentic tools that work on our behalf. But in some regard, it's a little bit like the dog chasing the car who finally catches the car, and then it's like, okay, now what?
And so I'd love to hear Money. Yeah. Well, yeah, yeah, I know. I've seen Money. All sorts of build me hey, CalladBot build me a b to b sales business that generates $50,000 a month, making no mistakes. Yield farming. Yeah. Well, that's the thing is I'd love to hear from from you guys on, like, any ways that you are currently using this or have plans to use it that you're willing to share.
Scott (Better Brand)The the agent that I'm building now, the it's done when it is able to yield farm and produce more money than it started with.
Bowen BelmerSo how how are you defining that? Like, are you testing it against because I saw one guy that had, like, 50 agents set up that were trading against each other, and he's eliminating, like, the bottom five or something like this every day.
Scott (Better Brand)I'm going for penny one. I and I'm so that I can figure out how to get there so I can put multiples in. But the that's the goal in setting up this agent from the get go and using a prompt that was worth a damn to get there. And I had to swing at this two or three times to do that, but it now knows that its job is to do what I can't
Bowen Belmerand to make money. Is it is it so you have it connected? Like, are you able to share like what protocols it's talking to right now? So presumably it has like an Ethereum wallet or something like this, and you're having it cross reference just all of the dApps that it can find?
Scott (Better Brand)Yes. I it hasn't done a whole lot of the financial stuff yet. It's just getting it's literally wobbly coat legs under under using the wallet. But the the trick to the wallet was putting it on the Mac key chain and never giving the the private key to the to anything, basically. It exists as a variable. Mhmm. So that that was the trick of getting that set up.
Bowen BelmerInteresting. Okay. Yeah. I mean, I think trading in general is just gonna completely change in the way that the markets work when you don't have people, interacting with them. It's like who can have the smartest agent, which is very, very different. Kyle, I didn't mean to cut you off. I think you're about to say something.
Kyle (Bow Tied Blue Fin)Yeah. I was gonna say something, but it the smartest agent, and it it it'll also be interesting, Bo, if you remember back to the Internet latency wars of, I guess, ten years ago or or so of people creating data centers closer and closer to New York Stock Exchange and so on. It it's it will come down to intelligence and speed. So that's an an interesting point, when trading against
Polymarket or trading on very small fast signals. So it it'll be interesting to see there. I think that the comment and this pairs well with privacy as well on you know, David spoke a bit about basic users, you know, grabbing a Mac Mini or an old computer and running it. I think there's kind of a sliding scale of that people need to be thinking about as far as what am I giving this access to and what associated security do I need. And it's kind of more access means more security. So
when you're giving private keys, for example, you may want a more hardened security solution than if you're just giving it access to a sandbox that it could only talk to itself. One thing that I've implemented recently, and I know many other people on this call have, is, like, a one password integration. So you pretty much just create a separate account or a separate vault, store everything in one password, and then have it make the calls to you know, you can there's a one password skill or or you just integrate one password into your CloudBot, and all of those secrets don't live, you know, on disk. They they are referenced when they're needed and then discarded,
you know, when when they're not being used. So or or in cache. But, it's very you know, different needs for different use cases. So I'd I'd recommend always thinking about handling things and thinking about the blast radius when you're you're doing things, especially with private key. Yeah. And, Kyle, I know that you went the opposite route of the Mac Mini
Bowen Belmerstraight to a VPS. Can you talk about your decision behind Macs? I know when we were talking about it earlier, ahead of this call, I had made the joke that I I felt like the Mac Mini was the easier route from for somebody who's less technically inclined, and you had the opposite take. So I'm curious, to hear more about that.
Kyle (Bow Tied Blue Fin)I I think the the opposite take on on what we were talking about before, and I'll just bring everyone up to speed, is the the comment was I'm I'm concerned about security, so I'm I'm scared of creating a VPS that doesn't have the right security. And and my take was, I'm scared about giving access to my home network and just knowing that I'm not a security expert by any means, you know, for,
you know, using OpenCloud in terms of blocking the right ports on my home network, making sure that things can't talk to each other, and so on. So I was like, let me just go completely, you know, separate put it on a VPS and and then be able to control the only people that can talk to the VPS. So maybe, you know, there there's a couple funny memes where, you know, left curve, buy Mac mini, right curve, just buy a damn Mac mini, mid curve. I'm gonna create a VPS with a tunnel and and all this stuff. So I might be very mid curve here, but I I saw the opportunity of spinning up a VPS for, you know, $5 a month or whatever it is and
just completely locking down, you know, where it could communicate. And you can SSH into the instance, You know, Claude or or or any AI understands how to connect to a VPS, and it's able to kind of walk you through that process, as well. So I took the VPS route personally. I understand the simplicity of Mac Mini. I think it's it's the same sliding scale on on what you're looking to do and and why you're looking to do it. Hey, Bo. I I think another use case that might be interesting to speak about
David Johnstonis the one of how this displaces software as a service. Right? I've been paying, for example, for a long time for a service called Strikingly, and it creates websites and hosts them for me. And what I realized is my agent can not only code those websites, but I can host them on GitHub. And GitHub offers GitHub pages because these are just static landing pages and offers that service for free. Right? So if I code that up and host on GitHub,
I can delete that $100 a month or whatever I've been spending on strikingly for a decade plus and just move over to effectively building and and hosting my own websites without the cost. And I think we're gonna find a lot of software that we're paying for today that all of a sudden, like, hey. I can spin that up and and have an alternative of that, and it's personalized to me instead of, you know, putting my credit card into the and then into the SaaS machine. I think this is a really fundamental
shift. Big report came out of Goldman Sachs where they're expecting a huge decline in SaaS revenues and a huge increase the next four years in AgenTic revenues and by 2030 for them to actually flip.
Bowen BelmerYeah. I mean, I don't doubt that. I think SaaS in general is really in trouble. I don't know. I'm sure pretty much everyone on this call has either, saw or read the something big is happening article that was shared by Matt Schumer yesterday, or was it yesterday? Two days ago now. Had, like, 75,000,000 impressions, which is outrageous. Although there is quite a bit of speculation that AI was actually used to write it, which is pretty ironic and funny. But it talks about that basically is is, you know, we have really officially crossed this chasm now where
I mean, everyone knows that AI is ultimately going to change the world of work, whether that be replace jobs or at least the very least change what jobs look like. We are certainly there, especially with things like software as a service. My recommendation, not not that anyone should specifically be taking advice from me on this because they shouldn't, and this is not advice. But with that disclaimer out of the way, my recommendation would be taking what you already do for work and trying to figure out a way to automate yourself,
essentially, so that you can at least interact with the thing that may be there to try to knock on your door to take your job at some point. You know, it would be like if you were doing accounting prior to the computer and you were the first person who really understood Microsoft Excel at your firm, you know, it's you're really insulating yourself with knowledge in a workspace that you already have and just taking that step forward and, okay, well, how do I how do I now be the one with the keys to the castle based on the new tooling?
Scott (Better Brand)Do what scares you the most.
Bowen BelmerExactly right. Kyle, did you have something? Oh, might have lost him.
Scott (Better Brand)One of the other incredible things is the the hello world moment when this thing messages you back or wishes you good night and you aren't really sure if it was the person that just messaged you. It it it's these are phenomenal firsts that you get with this thing.
Bowen BelmerYeah. Absolutely. I was actually, I should know his name. I'm so sorry. Let me just pull it up so I don't get this wrong. Peter Steinberger, so who who actually made OpenClaw or at least was, you know, credited as the founder, and I believe it was mostly self developed, and then now, of course, there's been tons of people who've jumped in to help out. He actually was just on Lex Friedman's podcast. The interview came out about fifteen hours ago and is about three hours and fifteen minutes long, so I have not had the chance to listen to the whole thing yet. But I do suggest people check it out to hear more about the backstory. But he actually talks about that in the early part of the interview,
how there was something that he had wanted it to do. Oh, it was, text to speech or or speech to text. His the ability for him to just talk to it. Apparently, it, like, taught itself, basically, and he wasn't necessarily anticipating that. And then all of a sudden, it it could speak to him, and he or he could speak to it rather. And that was kinda like a crazy hello world type of moment. And I suspect that we're gonna see many more of these in the not so distant future.
Scott (Better Brand)Yeah.
David JohnstonI mean This is your first robot. Yeah. Right here. Pee people are giving it a name. Right? They they they've got, you know, Bob, their agent or whatever they they named it. And when they send me stuff, they're like, oh, here's what Bob figured out. Right? So it's got a name. It's got a persona. It can talk to you. It can get on calls with you. I called it on signal and it called back. That that is a wild wild moment. Though I will say if you try to run it on too small a model, right, it it's like it's like, hey. Where'd you go? Like, it it the the consciousness, the the the personhood disappears as soon as you try to run it on, like, Llama 3.3.
Like, that's not gonna cut it. Right? There is definitely some threshold. That threshold is probably Kimmy 2.5. I guess the new GLM model is now even passing Kimmy. So there's a line where, like, you you can feel it when, like, it's getting things done quickly, and it can articulate, and it it has some real intelligence behind it. So, one of the things that I built into to Everclaw is a router. And so it will now intelligently route
tasks that are, oh, just writing or something versus complex stuff like coding to different models. Right? So if you set it up, you know, with with, multiple models, all of a sudden, now it can also save you on inference cost. If it's not throwing, you know, quad 4.6 or Kimi 2.5 at every problem, right, it can do some more intelligent routing. But, yeah, it's it's just wild.
Bowen BelmerI'd love to hear. Do you wanna call up maybe any community members to give their take on this? I'm sure tons of people have perspectives on this given that it's, you know, really just taken everything by storm over the last couple weeks.
David JohnstonYeah. I mean, that'd be great. I see, Brian's online, Matt, and others. Especially if somebody hasn't set it up yet, I'd love to hear, like, what people are concerned about or are hearing about it or want to know when they do their setup.
Scott (Better Brand)I'm also finishing a guide for how to set up your Mac Mini for it specifically, including the the dongles for the headless and settings you want to consider for having it survive a restart and how it accesses resources.
Bowen BelmerI will personally use that, Scott.
Scott (Better Brand)Awesome. Just message me if you want. Copy. Yeah. That'd be great. It was funny.
David JohnstonAll of these OpenClaw events are starting to pop up all over the world. So I saw there was one last night or the night before in Vienna, and there's one tonight in Austin. They're calling it the Clawston event. And, yeah, gone soon. I I'll be curious to see how many people show up, but it's tonight from six to 9PM. So if you're in Austin, just go to Clawston, clawstin.com. It'll show you,
the event, but it should be, pretty wild. All the AI people I know are are coming out, this evening for that. So I'll be there. I'll talk a little bit of at the event about Morpheus and how people can use Morpheus inference so their open claw never runs out of inference, that they should own their own inference. And all you gotta do is buy some more to to do it. So it should be a fun event.
Alan (Nomadic Rogues)How long before that event goes from in real life to across agents?
David JohnstonYeah. I mean, that's one of the things we were even talking about last year is it we came to the conclusion, oh, I'm just gonna want my agent to talk to somebody else's agent. And that's not really what Maltbook is. Maltbook's a little confusing because it sounds like Facebook, but it's not Facebook. It's more like Reddit for agents. So somebody, you know, can have their agent go read all the posts and upvote them or downvote them or create a new post. But they're not really talking to each other as much as they're posting and ranking
useful information. But what I would love, and I expect to emerge very soon, is a way for my agent to talk to your agent, Thomas. Like, hey, if we're both working on the same open source project, I might be way more willing to, you know, let you have broad permissions to talk to my agent in the context of, like, developing open source for Morpheus. Right? And and effectively, all things move at the speed of trust,
and humans already have a good sense of who they trust with what. Right? You've got a different trust level with your spouse, with your kids, with your family, with your friends, like, stuff that you would never put, you know, on the public Internet. Right? And so, yeah, we'll we'll we'll see how quickly that develops, but I expect that something that's gonna pop out. And then we're gonna have, you know, 8,000,000,000 humans with each with an agent all talking to each other twenty four seven. Right? And and that's gonna be a beautiful moment because
then it really goes exponential. Because right now, we're we're sort of still doing this manually. People create a skill. They post it on x or they put it up on on ClawHub. Right? And people start to these, but they don't have a good sense for how trustworthy they are. So yeah.
Bowen BelmerDefinitely. It's just unironically so bullish for in person meetings.
David JohnstonYep.
Bowen BelmerBring bring back social gatherings.
David JohnstonYeah. That that'll that'll be a big part of it. But maybe one more thing to highlight on the the package of skills in in Everclaw is I had the agent grab both x 402 and 8004. So if there are interesting agents, for my agent to pay because it has a wallet, now with x four 02, it has an easy way to do that. Right? And I again, that goes with the whole trend of getting away from SaaS. I'd rather just pay, you know, 25¢
for my agent to go get something instead of sign up for some $10.20 bucks a month service. Right? So I think this is where x x four zero two really gets legs. The second thing is everybody should know trustless agents. The the Gentec registries are live on Ethereum and Base and a lot of the other chains now. And so you can finally have reputation for every agent. Right? It validates not only the identity, not only what it does, but actually gives you a reputation score.
So now I've set it up so your agent can, access the 8004 registry. Right? It could just go to, 8004 scan, see all of the agents. So when it's looking at a new agent to use, you get a sense for whether it's actually trustworthy. Right? And both of those tools are super early, but it's it's really interesting that we now have all the primitives. Like, you couldn't have built a lot of this stuff a year ago or even six months ago. I think eight zero zero four went live last week.
Right? So just in time for the agents, we're gonna find out there's a lot of malicious stuff out there, and they need to protect themselves, and and reputation is is part of that solution. But it's also worth mentioning the clause very much in beta. Right? So give me feedback. Let me know what doesn't work. There's probably a ton of ways it needs to improve, but that's the whole idea of open source. Right? So I've put all the Everclaw skills out in MIT open source license.
So fork it, use it, test it, do whatever you want. But, you know, I I I'm very excited that we now have this flywheel going where we can keep improving these tools and agents and tying in Morpheus and everything that we've built the last couple years. It's just I like I'm I'm I'm spending every extra minute, you know, playing with this because it's it's so fun.
Brian BourgerieK. So I got two questions for the group, if that's alright. First off, in your experience for those who have launched an agent already, how close are you to actually or how how far away are we at this point from you feeling comfortable actually giving it access to most of your stuff? Like, if you're wanting a true executive assistant or chief of staff or or something like that. 100%. How how long how far away are you be, or are you already all in on that?
David JohnstonI I have not taken that dive yet. And and I've been perfectly happy to spin it up dedicated accounts. And so then instead of acting, you know, automatically for me, it sort of gives me one layer of like, it can send the email or it can make a poll request on GitHub, but I can approve it. But I think I'm gonna get there fairly soon that I'll give it more and more access. And then what do you think, Scott?
Scott (Better Brand)I wouldn't want this thing acting as me as much as someone that works with me so that there is some sort of distinction with where information comes and goes. So like you said, giving it its own, I guess, ability to purchase, communicate, and and reason is just a matter of you now figuring out what its guidelines need to be as opposed to what it has access to, which makes it easier.
David JohnstonYeah. And because and because of the security stuff, you know, even for its own access, I've kinda put it in, like, read only mode where it's, you know, reading MaltBook, but it's not posting the MaltBook. It's not talking with any MaltBook. It's just using it as a source of information. And and today, I've only white listed myself to interact with with the agent. But all I'd have to do is update a setting, and my agent could be talking to Thomas. But I probably need to
spend more time thinking of how I set the settings for, like, what can be accessed by who. Right? And maybe that means, like, spinning up a different agent that talks to friends or a different agent that only talks to, you know, anybody on x that reaches out. Right? That those are probably different personas, and I might have three or four of those for friends, colleagues, public, family, whatever. Right? But I think I think I think that's gonna get there pretty quickly, quickly, and a lot of people just YOLO it into giving it everything.
And that's where all the crazy stories have come from of, like, oh, I did this, you know, or it spent all my money or it send all these crazy emails to my boss or you know? I mean, those are fun those are fun for the lulls on on the Internet. But, like, I'll take a little more cautious approach, but I don't think it'll take me too long to get there. So I think we'll figure out how to do it.
Alan (Nomadic Rogues)Well, one step that you may wanna consider is having having one agent that filters that communication or checks the communication, checks it for tone, checks it for consistency around the way you communicate so that instead of creating multiple personas behind you, you essentially have a fact checker, if you will, right, or a tone checker before anything goes out. So the agent the agents do their thing, but before
everything goes out, you have this sort of firewall agent that will tone check everything before it before it exits the system. And that's that's something I've been using for all types of content creation was it like, the agents do their work. They create whatever needs to be created, but then it gets funneled to a fact checker and a tone checker. And then I only look at what comes out of this.
Scott (Better Brand)I want mine to be wrong all the time. That's gonna be the entertaining part is to just see what it figures out and what decides to do. So I I I want a balance of that.
Thomas BorrelWell, it's it's funny. Listening to some of this conversation, what pops into my head is raising a kid. Right? You don't all of a sudden give the keys to the car to your one year old, give them the credit card and let them have free access to the house and everything else. You've got to grow along with it. So you give them their own identity, you give them the playpen to play in, ring fenced, firewalled, you know, your own Mac Mini. And then as it grows and learns and as you trust it, you give them
more and more responsibilities and you're allowed to develop as their own identity. And I think that's the parallel that pops up into my mind quite a bit is you're you're and I hate to anthropomorphize too much, we it's such a struggle not to, especially when you call them Bob or any other name, but this is growing in AI, growing in artificial intelligence. And so you so I see and I fear, especially as a security conscious guy, the they could say just wing it out there and all of a sudden, you know, next week, why poor? Well, you you gave it your bank account said you want to be richer so it it made a bad bet. Whoops.
Yeah. So, I think if you I think if we treat it like the growing entity that it is, you'll be safer, you'll grow, we'll have to mature along with it. But that's that's the thing that rings in my head as as much as I think about it as a growing child. You're not gonna give the the five year old the keys to the Porsche just yet. Can prove themselves a little bit.
Scott (Better Brand)An allowance. I like that. Yeah. He can lose its money, not my money.
Bowen BelmerWell, is huge for those of us that do not have children. I'm gonna let my parents know that, hey, you know, you don't need grandkids. We have we have Bob, the agent. Here's their signal number. They're sure they'll love that. Digital grandkids. Also Yeah. Also, I wanna I wanna put it out there that I specifically told it not to lose any money. So you know? That's kind of it.
Scott (Better Brand)It it was hesitating a lot.
Brian BourgerieI tell my kids not to do a lot of things that they do. So
Thomas BorrelDon't don't lose any money so they could just convert it all to Bitcoin and and black hole it. That's right. I didn't lose it. It's protected forever. You just can't get to it.
Bowen BelmerNo one can. Yeah. I can foresee a future though where you kind of have, like like, your household agent and then delegate agents below it where you're basically allowing certain things for, yeah, like your your kids, your, you know, I I don't know, whoever might need access to that information, and it's kinda making those judgment calls for you based on the parameters. That's interesting.
Brian BourgerieWell, actually, along oh, along the lines of multiple agents, that was another question I had for the group because I've been making my plans for having one of the venture studios I'm running just be completely just me and all agents for, like, everything else. And I'm looking for the best frameworks and structures. And, like, one, I've I'm sure some of you guys follow, like, I I'm I've been following Austin with Gauntlet AI. Like, he spun up his
Kelly Claude single, but before he knew it, he has an entire factory, and they're pumping out, like, what, 16 projects a day or something like that. And I was just curious if any of you have thought about the growing your agent family, what that looks like, if you found good frameworks for having your swarms or your entire leadership executive team or whatever that flavor is that you're you're looking at, or if you've just been fixated on the the solo agent?
Scott (Better Brand)When I get one right, that's that's the one I'll try and do more than once. But, yeah, no. I've been going back and forth on that. It's I'd love to know what that guy's token use is.
Brian BourgerieI'm sure it's insane.
Scott (Better Brand)Yeah.
Brian BourgerieI don't know. Has anyone else thought about or have, like, a swarm strategy or a team of agent strategy yet, or is it too early for most people to be thinking on that front?
Alan (Nomadic Rogues)I think it'll I think it'll get there. Go ahead, Thomas. Sorry. No. I I have something along those lines, but I control the communication between them. So I still have sort of a a human in the middle between what one group of agents builds before it passes to the next level for, like, execution or further analysis or whatever the work is. I'd I'd like to take a look at it to make sure that you don't have a drift too much of a drift as it
as the information passes from one agent to another to another to another. So it's not fully automated yet, but, yeah, I've got a bunch of different functions, if you will, served by a variety of agents, and the output of one goes to others for them to do their work. It works well. I can't wait to make it fully automated, but it it it'll be a question of time. Right? I think that that trust needs to be that trust needs to be built,
and and that's where the this construct of a gatekeeper and and fact checker and tone verifier is is something that I've injected in the system to slowly remove myself from from
Brian Bourgeriethat chain. And those those aren't all set up through OpenCLaw, though. Right? Those are that's a separate
Alan (Nomadic Rogues)framework or structure? Or Correct. Yeah. Yeah. Correct. Right now, they're not set up through they're not set up through OpenCLaw, but they will they will soon. They will soon.
David JohnstonYeah. The speed at which this is moving is just incredible. Right? You see folks like Coinbase over the last couple days shipping tools specifically to Clodhub so that people can use the Coinbase wallet for payments, and it's structured as a skill that you can give to your agent. Right? And so it's so cool to see sort of everybody leaning in. Stripe also announced that they've integrated x four zero two. So it seems like most of the major fintechs
have now gotten to the point where they're using the on chain payments, and Coinbase is leading the way with the agentic payments. So, you know, I I'm just happy that all these pieces are coming together. Right? If this had hit a year ago, we wouldn't have had the payment rails. We wouldn't have had the trust rails in place, and you would have got sucked into a much more SaaS like model for all of this stuff. I mean, consider that OpenClaw is open source,
and everybody's running their own. And they can run it locally. And if they need more powerful model, they can tap Morpheus, and they can actually own their inference. Like, this is what we've been dreaming about. Right? This is you you know, know, OpenClaw is the one, right, that Morpheus has been searching for the last couple years. And I'm so happy that this is the way it's playing out, and we we can get away from these big corporates that have had a a cloud agent LARPing as your personal AI.
Right? This can actually be a real personal AI.
Scott (Better Brand)It's your DIY first robot, and it's yeah. You're you're building it yourself. It it'll be the story that you tell a lot of people as you become,
Bowen Belmeryou know, the old guy making obnoxious stories. Yeah. Do you think when we get to hardware so, like, you know, we saw Homebot and these other things being teased late last year. Do you think when we get into hardware, one of the earliest versions will be somebody making, like, a a Homebot type thing that I can just plug my Claude bot into? Do you think a Claude bot is gonna be relevant four months from now, or is it is this just the first of many
Scott (Better Brand)for things like many. Yeah. Definitely first. I mean, we were all talking about Claude code two weeks ago, and that was the greatest thing ever. So but the I can see a world where a tool is an arm and a skill is a finger and programming it just like that and going out and buying an arm with a 400 pound PSI and then just basically setting the skill at a certain level and boom, there's your arm. And just putting it together just like that with like Legos.
David JohnstonBo, I would make I would make the argument that I think OpenClaw is going to stick around for the long term because it's a framework. Right? It's it's the kernel at the heart. Right? It's managing your files. It's managing your connection to models. So whatever model comes out tomorrow, I'm just gonna have my my OpenClaw agent use it. Right? So whatever the next great model is fifteen minutes from now, great. Well, that just makes my agent more powerful. But what I think it's sticky.
Once you've spent a few hours giving this thing accounts, giving it, you know, access to to things once you once you've defined its user persona and its soul, like, I don't know if I need to do that again. Right? I I might copy my OpenCLaw and put it on a second machine, or I may create different personas, but I think that this framework is really sticky. Right? And as everybody integrates with it, it makes it more sticky. Right? Because you've got a network effect growing around it
that, you know, is is hard for somebody to displace. And, like, what would what would it look like for somebody to displace it? It also has to be open source. It also has to be something that can run locally. Right? And and for for retail people, I don't know if they need special hardware. The Mac mini is a decent solution. I would say for, you know, professionals that, you know, wanna use this at a at at, like, a, you know, work level, yeah, there might be something that that emerges on on the hardware side.
But I think for a lot of people, this is a huge boon for for Apple
Bowen Belmerat you know, and and it's sort of being at the heart of a lot of their their personal AI stuff. But, you know, I don't think it's a fad. Like, everybody, I think, a week or two ago was thinking of this as a fad. I think this is a a real breakthrough. Can can we talk about, I guess, what the moat is, David? I mean, you kinda just touched on it, but, like, I'm I'm in the headspace right now, I think, of somebody who would look at Bitcoin and be like, well, why can't I just copy paste Bitcoin and make Bitcoin too? And, like, the answer is you can, obviously, literally do that, but but the network effect is just so large. Is it basically the idea that, like, you think the network effect for this will be will be kind of the same thing? I guess, is there a tech like, a technical moat? Because for Bitcoin, it would be, you know, the proof of work. Like like this, the the the network effect of having
a high difficulty rate and a large amount of people mining.
David JohnstonSo in this case, the proof of work is everybody running their own OpenCLO and making it better. Right? So what I what occurred to me last night is we are the singularity. We're building it. We're shipping it, and we're sharing it with each other. Anybody that comes, you know, a few months from now and tries to do OpenClaw two point o, like, one, try to break through the noise and people hear about it. Right? That's part of the whole, like, magic lightning in the bottle is it went viral and all these crazy stories of the things people were doing with it, you know, got it into the public consciousness.
Right? And so and anything that somebody comes later and they're like, oh, well, I did this. Great. Okay. Well, you know, I can have my agent add that in. Right? But what what's what's interesting here is there's this whole meme about recursive self improvement. And the AI labs have said very clearly, like, they think that'll land in the next twelve months. Like, Quad is already basically doing it, which is the Quad version you use today was programmed by the Quad version from yesterday.
Right? There are almost no human coding that's going on in Quad anymore. It's entirely, you know, the agents doing all the code. Right? And so that's an interesting loop, but I would argue this skill this recursive skill improvement is more interesting because it's organic. It's driven by people's real use cases, and the intelligence of a billion people out there is gonna be way more than the, you know, 5,000 coders at Kwan can ever imagine about the use cases that people want to do. Right? So I think this loop that we're now in
is the singularity. It's not waiting around for a super smart model. It's everybody having an agent and using it to make all of their agents better. This is the open source singularity that we've been hoping for. So it's for me, it's extremely encouraging because we've been waiting. We've known this moment was coming, and it feels like this this is it.
Bowen BelmerI'm equal parts excited and terrified.
David JohnstonWell, it it speaks to why you wanna spend the time to set it up right. Well and and that's the difference. It's like the person with their own hardware and who does it right and runs it because there's gonna be a lot of people that come out and try to sell open clause of service. And, you know, just pay them, you know, x dollars a month, and they'll run all the models and all the things for you. That's not real personal AI. Right? But we we have the opportunity, and I think it's incumbent upon us right now to get our friends, to get our family up the learning curve on this. Like, there's rumors. Last week, the rumor was
OpenClaw or at least Peter, the founder, was gonna get a giant check from Meta. This week, the rumor is he's gonna get a giant check from OpenAI. Now the code's all open source, so I, one, wanna have my thing before some tech company tries to shove their, you know, arm into it. Right? But two, you know, this this is an important moment. And if and the further this can get, you know, purely as open source, I think the better off we'll be. So, you know, this is the time to to bring your friends and family off the up up the learning curve. Right? Because they're gonna start asking about this because they're gonna hear about it, and you can help them set it up in a way that's actually gonna benefit them.
Bowen BelmerOh, sorry. I was talking on mute. Classic. Yeah. I I think that we're gonna look back on this time as, like, a we we will remember this moment in time as, like, the the first time that it crossed the chasm of, like, in the real zeitgeist of people having personal agents. You know, it's a thing that we were talking about for the past two years or maybe slightly more than that, and it's now a reality. And I, for one, cannot wait to see how people use them to do really awesome things, and sometimes not so awesome things.
But we're kinda right at the top of the hour. I mean, I definitely wanna leave some space open for anyone who wants to jump up and either, you know, add commentary or answer a question. Otherwise, I think this was great. And, yeah, I really look forward to seeing what happens with Everclaw. David, thank you for your efforts on that. I know I personally will be a user. I will be running to buy a new Mac mini this week.
David JohnstonAwesome. Good stuff. Now I'm glad to see Morpheus at the heart of this and putting to use all the hard work of Alan and Kyle and everybody who's been building this last couple years in the ecosystem. Now now is the time to give people the ability to own the brain that is powering their their agents. Right? So ownership, I think, is is important. Don't don't rent your intelligence from a big corporate. So, yeah, that'll be a big difference, I think, for people in the in the future. So really excited to see that. 100%.
Scott (Better Brand)It's also a great example of how to integrate Morpheus into something very popular.
Bowen BelmerYeah. Well, I mean, that's the thing. It highlights the, like, agnostic use case of the protocol. Right? Like, it it can be plugged into anything that wants it, ultimately. This is this is a perfect example of that.
Scott (Better Brand)Yeah. But we should not glance over the use case of this being a self sustaining agent living on its own hardware with its own money and not hiding under the bed because of that.
Bowen Belmer100%.
David Johnston100%.
Bowen BelmerAlright, guys. Well, thank you so much. I appreciate everybody as always. Whether you are a builder, community member, somebody just checking out what we're doing, definitely get involved. Morpheus has a million different ways to contribute, whether it's, you know, literally shipping code, whether it's building something on top of the protocol, whether it's talking about it or jumping into some of the social groups on things like Discord, get involved. You know, there are so many great people on this project and is
Scott (Better Brand)why it persists and will continue to do so, you know, long into the future. Thanks, everybody. Yeah. And if you want a guide on how to set up that Mac Mini, reach out. I'll have it tonight. Awesome. Thank you.
Zion, hear me! For Season 2 of the Mere Morpheus podcast I'm going to be uploading the Morpheus builder weekly updates. My reason for doing this is to make it more accessible for those who want an update directly via...
Zion, hear me! For Season 2 of the Mere Morpheus podcast I'm going to be uploading the Morpheus builder weekly updates. My reason for doing this is to make it more accessible for those who want an update directly via...
Zion, hear me! For Season 2 of the Mere Morpheus podcast I'm going to be uploading the Morpheus builder weekly updates. My reason for doing this is to make it more accessible for those who want an update directly via...
Zion, hear me! For Season 2 of the Mere Morpheus podcast I'm going to be uploading the Morpheus builder weekly updates. My reason for doing this is to make it more accessible for those who want an update directly via...
Zion, hear me! For Season 2 of the Mere Morpheus podcast I'm going to be uploading the Morpheus builder weekly updates. My reason for doing this is to make it more accessible for those who want an update directly via...