Transcript: Claude Code + Paperclip Just Destroyed OpenClaw

Watch video

Nate Herk | AI Automation20m 33sTranscript ✅Added May 7, 11:52 am GMT+8

Source video ID: HJ-dwefABss

Transcript

0:00 — What you’re looking at right here is my mission control center. You can see I’ve got seven agents enabled. I’ve got five tasks in progress. You can see what all of them are doing right here. We can see all of the recent activity and all the tasks. And then in my inbox, I can see what is needed to be approved by me or what I have to look at. Down here you can see the seven different agents we have and you can see what they’re actually working on. So right now the AIS marketer is running. And for each of our different agents, we can control their instructions, their skills, their configuration, everything. So now in a matter of 30 minutes of me coming in here acting as a board, which means I’m
0:30 — giving them high-level goals, metrics, you know, things that I want to achieve, but I’m not actually sitting in the business being an operator. I have this entire company now. I have a CEO, I have a social media agent, I have a marketer that manages a copywriter, a strategist, a designer, and a researcher. And this is all thanks to a combination of Cloud Code and this tool right here, which is called Paperclip. So by the end of this video, you guys are going to see exactly what this tool is, how it works, and you’re going to be able to either set up a brand new company right now with no human employees, just you, or you’re going to be able to build a team around maybe of your business. So for example,
1:01 — right now with my AI business and AI Automation Society, I’m not going to come in here and automate everything and go fire all my employees. But what I did here is I can automate a portion of it and then slowly just kind of like chip away more and more, expand this organization in here, and maybe grab a different sub unit for my business and try to make that completely AI first. You can see they’re sending me notifications down here. I can come in here and I can create new issues whenever I want. And I can say, “Hey, this is for the AIS designer and this is for the LinkedIn content strategy project. Here’s what I need you to do. Here’s the high-level goal. Work with
1:32 — the team and just figure it out.” So if this sounds like something that’s interesting to you, let’s not waste any time and get straight into the video. So the tool that we’re talking about today is called Paperclip. It is completely free. It’s an open-source orchestration for zero human companies where you can literally have your AI employees hire other ones, set goals, automate the entire business. You can see on the GitHub that they just hit over 36,000 stars, and this has only been out for a few weeks. You can bring in your own agents. So if you’ve already set up some custom, you know, Open Claw or Cloud Code things, you can bring those in. You set goals. They’re a heartbeat, so it
2:03 — can actually just be waking up and checking things all the time. You can set a specific budget per agent. So if I come to the dashboard and I go to the, you know, designer for example, I can come in here and just make sure that the designer is only using a certain amount of my money per month. Which means every single one of our agents, we can see exactly how much money they’re actually spending. Now the reason you guys are seeing zero here and on my dashboard is because I’m using my subscription. So if you’re using something that is actually paper token, that’s where you get all of the spend analytics. You can run multiple companies inside of Paperclip.
2:33 — So if you see on the dashboard over here, I’ve got this AIS, but I can go down here and just add a new company and add a new company and just keep doing different things. And so that’s where maybe if you wanted to start up a new idea, then you would start a company for that. But if you wanted to integrate some different like teams into an existing business, then maybe you could just do different departments in here and completely automate those. So for example, with this company, this is part of AIS, and the overall goal I set here is, you know, we have a community called AI Automation Society. It’s got over 300,000 but we want to scale the content efforts around the community and scale, you
3:04 — know, like the brand around this that doesn’t rely on me and my personal brand. So that’s what I’m going to be doing here in Paperclip. It’s got a ticketing system, so everything is logged, every conversation, every issue. If I go back to my dashboard here and I go down here, we can see all the activity and all the tasks. And I can open these up. I can leave comments and I can see what they’re thinking and what they’re actually doing in here. And in this specific one, you can see what it was doing is it was creating some carousels. This was its first try, so I’m going to have to go back, but very cool. And like I said, you can use this with Open Claw, Claude, Codex, Cursor,
3:34 — whatever you want. In today’s example, we’re going to be talking about using it with Cloud Code. So I know I was just clicking around and I showed you guys a ton of stuff. That’s just to sort of get you in the mindset of what we’re going to talk about today and how it works. But what I’m going to do is set up an entire new company from scratch with you guys, so you can see exactly what it would look like on your end, what you need to click, and what you need to know. So before we get into that, let me just do a quick explanation on Paperclip. Also guys, real quick, I threw together this 12-page document about, you know, what Paperclip is and walking through how you should get set up and what you can do with Cloud Code. So if you want to access this for
4:04 — completely free, you can do so by joining my free school community. The link for that is down in the description. But let’s get back to the video. So this launched in early March, so about 3 weeks ago, and it’s already trending a lot on GitHub as you can see with the explosive growth of the stars. Now the thing that I think is interesting is why did this blow up? Because there’s been lots of things going on, right? We’ve had like Open Claw, which took over. And then we had Cloud Code start to work in things like remote control and channels. And this idea of like being able to work more with your things from anywhere and have them be more productive and proactive. But one of the founders of this, Dota,
4:35 — he basically explained a pain point, which was every day he would have 20 different terminals running, 20 different Cloud Code sessions running. And I’ve definitely been there. I’m sure you guys have been there. And then what happens is, yeah, you have all this productivity, but then when you come back, you’re like, “Okay, I forgot which one’s doing what and I forgot like, what I asked and I forgot like, you know, I don’t I just don’t have as much visibility as I’d like.” So it seems like Paperclip is solving a major pain point of figuring out all of that orchestration and keeping everything kind of synchronized in one place with different sessions being able to actually understand the overall goal
5:05 — that we’re working towards. But anyways, it is fully open source. It actually has MIT license. And then here’s a little bit about how it works. Right now, when you just go ahead and launch it up, you’re going to launch it from a terminal, and then it’s going to open up, you know, obviously this pretty dashboard. This by default will be in a local host, so it only works on this machine. But what you could do is you could set it up on a VPS or something else, so you could access this from anywhere. But anyways, what you’re going to do to get started is you’re going to either going to go to the GitHub or you’re going to go to right here, paperclip.ing. I will put a link in the description. And you literally just
5:35 — click on get started. And what happens is it brings you down here to this command, and that’s all you have to run in your terminal. So if I copy this and I go to my terminal, and don’t worry if you don’t want to be in a terminal, you don’t have to. You just paste this in, and then what it’s going to do is it’s going to bring up the actual server. As you can see, it just opened up my computer, and it would bring you to this local host where you would now onboard. And because I’ve already onboarded, I don’t have to go through that. But basically, it would look like this. It would say, “Okay, cool. Name your company.” So let me set up some example stuff here. So we’re building a company called Proofshot. The mission and goal
6:05 — is to give our customers a link and basically they’ll be able to record a testimonial, and then AI cleans it up and is able to let us embed that on our website or things like that. So this is just an example. And so let’s go ahead and get started with this company. We have to create our first agent. So this doesn’t have to be a CEO, but we’re going to go ahead and do a CEO. And once again, because this is running locally, if you already are authenticated into Cloud Code, you know, somewhere else on your terminal, then it will just be able to connect pretty much immediately. For the model, you have a choice. Right now, let’s just go ahead and go with Sonnet 4.6. And then you can just run a quick
6:36 — test to see if it can actually hit the CLI and respond. And you can see we just passed. So I’m going to go on to the next step, which is our first task. So this is kind of the default prompt. It’s going to say, “Hey, this is the first task. Hire your first engineer and create a hiring plan.” And this is obviously going to the CEO. We also have a quick description here. So this is where maybe you could add a little bit of background context about your business, and you could maybe even change this if you know that there’s a different initial project you want to work on. But for the sake of the demo, let’s just go ahead and let our CEO make a first hire. So we’re going to move on,
7:06 — and then we’re going to say, “Okay, this all looks good. Let’s go ahead and start this company.” So now in our Paperclip dashboard, we have AIS that we could go look at, and you know, things are running and working here, or we can see our new company, Proofshot, and we can get familiarized with this environment here. So what you’ll notice is that we have one agent running right now, and this is the CEO. So what I can do is I can either, you know, click right here on this task, which is number one, “Hire your first engineer.” And we can see it right here actually running. So this is basically the way that it’s running, you know, in Cloud Code when you talk to it. It’s creating a plan. It, you know,
7:37 — executes commands. It can do web fetches. It can do the things that Cloud Code can natively do. And then it’s basically going to give us a task in our inbox and it’s going to say, “Hey, approve this.” Or hey, you know, leave a comment here. And that’s how we’re able to manage it. Real quick also, while this is running, let’s go ahead to the actual agent. Because right now, you know, we only have one. We only have our CEO. Let’s take a look at what we’re actually looking at with our agent configuration. Right here, what we can do is we can assign a task or we can run a heartbeat. Now what exactly is a heartbeat? So native heartbeats is one of the things that made Open Claw blow up because it
8:08 — felt way more proactive. And basically what it means is it just keeps the agents running. So the agents basically wake up on a schedule, whether it’s every 4 hours, as you can see, every 8 hours, or every 12 hours. And when they wake up, it’s almost as if they were just born. So they wake up with fresh context and fresh memory, which is why it’s important that they have to check their work. They look at their tasks. They get familiarized with their environment first. And that’s how they’re able to just keep working around the clock. So you may be wondering, once they wake up, how do they know what to do? And that’s
8:38 — where you could come in here to the instructions. And what you’ll notice is that the CEO has four different files. It has an agents file, a heartbeat file, a soul file, and a tools file. So these files basically work together to make sure that every time the CEO wakes up, it knows what’s going on. So you can see here, you are the CEO. Here’s your directory, blah blah blah. Here’s how you use your memory and planning. Here’s some safety to And a quick explanation of these, the heartbeat is execution and extraction checklist. It runs on every heartbeat. The soul is who you are and how you should act. And the tools are
9:09 — tools that you have access to. And these files, of course, are going to evolve and grow over time. And you can also come in here and manually add things or delete things yourself. You can see the soul is a CEO persona. So you own the P&L. You default to action. You hold the long view while executing the near term. You protect focus hard. You can also customize the voice and tone and anything else that you want in a CEO. Now what else is cool is you may have just installed Paperclip and be a little confused about it. And there’s two things that make this a ton, ton better. The first one is the fact that natively
9:40 — all of these agents will understand Paperclip. They have all of these Paperclip skills already installed and already loaded up. So that when you tell it to do things, it knows exactly how Paperclip works and how it needs to actually do things. Now the other thing that I did, which is super, super helpful, some people might say it’s overkill, but I definitely don’t think so, is I set up my Cloud Code project to understand Paperclip and to help me get everything configured. I even went as far as to create a Paperclip project in my executive assistant here, so it understands this. Because here’s what it knows.
10:11 — It understands the full Paperclip architecture, the API, the heartbeat protocol, how we could eventually move this over to a VPS, it understands all of that. The agents I have in there, the secrets I have in there, how to query everything, the gotchas, and it can help me as I want to expand this with monitoring, with configuring, with adding secrets, new agents, like I said, planning the VPS migration. And this is basically just going to be like my partner in crime when I need help doing something with Paperclip. Because the way I feel about things is I don’t always know everything, right? And we’re human, so we forget things. But if you can set up a project where you give it
10:42 — the GitHub repo, it’s open source, right? So it can know everything about it. You give it the GitHub repo, you let it look at X to see how people are using it, you let it do its own research and find its own things, and it’s going to help you use tools so so much better. So that’s why I set up a Cloud Code project to help me out with my Paperclip projects in here. And as you use Paperclip more, you’ll get more used to it and you might not need that Cloud Code agent as much, but to get set up and to get oriented, I think it’s a great idea. We can continue on in the configuration, we could change the name, the title, we could also change who it
11:12 — reports to and what the capabilities are. And here’s where you could change the actual adapter. So per agent, you can have a different model, you can have a different provider, and that’s where it gets so so cool, because you can really customize this, you know, org chart to be specialized agents for specific tasks. Now there’s so many settings in here, of course, you can do things like letting them actually assign tasks or letting them actually create new agents. There’s a lot of things to play with, but we’re just going to keep it simple and I’m going to keep moving along here and building this company. So we can also see the runs that they’re doing, so right now there’s only been one and it
11:42 — looks like it’s actually sitting in our inbox waiting for us, so I get to that in a sec. And then of course you come to the budget and this is where you can actually enable a budget to keep things in check. So I’m going to go ahead and go back to the dashboard, see what’s going on, and we can see we have one pending approval. So I’ll click on this, it brings us to the approval box, and we can see what it wants to do is the CEO has requested to hire an engineer, a founding full stack engineer with capabilities of, you know, integrations, design, database, front end, and it sounds like for our kind of like SaaS
12:12 — product we’re building here, we’re building this company around, we need this. So I’m going to go ahead and approve that. And what that does is it creates a new linked issue. I’ll see if I click into this what’s going on. We can see that this is now running once again, and the CEO is getting this engineer set up. You’ll see if I come down here to the agents, we actually have an engineer now, and this engineer has been set up. Currently it just has a very minimal instructions, it doesn’t have much to do here. Really all it has is just the Paperclip skills and it’s sitting here and waiting for the CEO to build a plan and to assign some work to it. You can
12:43 — see the CEO just started up a bunch of tasks, and now the engineer has five things going on. So if I go back to the dashboard, we can see all of the stuff that is going on. I’m going to go ahead and go to the issues to see what tasks were made. We have milestone one, which is an MVP core back end, milestone two is an AI processing pipeline, front end recording UI, embeddable testimonial card, blah blah blah. But what I notice now is that we don’t have anyone kind of assuring the quality. And what I’ve noticed is that we’re not actually putting all of this in a project, and we probably want to cuz it’s going to be
13:13 — building code. So if I go to projects, we could actually sync this to a GitHub repo, so it’s optional. Let’s just pretend that we’re going to put this into a GitHub repo because, you know, we want to keep the code base there. But let’s just call this project initial MVP, and this is basically getting our first version of the website up and running so that we can continuously improve and refine. We can set the status, we can set the goal, and we can also set the date. We’re just going to go ahead and create this project. So now I can go ahead and create a new issue. I’m going to assign
13:43 — this to the CEO inside of our initial MVP project, and I’m going to call this migration and QA. So I’ve noticed that you started assigning some tasks to the founding engineer, which is great, but we don’t have anyone to do quality assurance. So we need you to go out and hire a QA agent that will make sure that everything we’re doing is quality, making sure that it’s safe, and making sure that it doesn’t have any bugs and things like that. The idea here is that you should continuously build and you keep making improvements and refinements
14:14 — until you come back to me and say, “Hey, the QA agent and the founding engineer are done.” So that was kind of a long goal, but you can see here we’re just acting as the board. We’re giving high-level instructions, we’re giving the outputs we want, and we’re letting the CEO figure out who it needs to hire or who needs to be, you know, like looped in on this plan in order to do this. All right, so while this is running, let’s talk about skills, because you saw that we could give each agent individual skills, and you also see that we have a company skills section. You can see we can paste in a certain path, a GitHub URL in order to
14:45 — add skills. So something that I probably want to install here is the front end design skill. So I’m just going to go ahead and copy the URL of the front end design skill repo. I’m going to go back into the dashboard here and I’m just going to paste in that skill and click add and see what happens. Boom, we have pulled in this front end design skill from GitHub right here. Now if you’ve never heard of skills.sh, it is kind of like a marketplace for skills. They’re all pretty much free on here and there’s tons and tons of different ones for different types of, you know, processes. So here you can see this was the front end design. You can see we
15:16 — have web design guidelines, so maybe we want to pull this one in as well. Now one thing to keep in mind is they do have security audits and things like this, but you still have to be careful and you still want to make sure that, you know, you you never ultimately know at the end of the day. So just be careful when you’re doing this, but if you wanted to install the skill, you would copy this right here, you would go back into your dashboard, and you would just paste in that path. And now we have added another skill to this project, which is the web design guidelines. So we have gotten this set up, right? We have these agents working, we can actually click into this one right now,
15:47 — and we can see what the CEO is doing. You can see the task is clear, I need to hire a QA, I need to migrate the tasks, and it’s gathering context, and we can watch it work. We can see the way it thinks, just like when you are inside of Cloud Code. It’s just a little bit different the way you actually interact. It’s much less like having a, you know, a long ongoing conversation, it’s more so creating issues, leaving comments, you know, waking them up on a heartbeat, things like that. So another thing that we haven’t talked about yet, which you probably noticed over here, is routines. And this is currently in beta, but basically we can use the create routine
16:19 — to define the first recurring workflow, which is awesome. So if I go ahead and I want to create a routine, let’s just say we wanted to do a, you know, security check and we wanted to do this, you know, every single night or something. We could assign this to someone, so maybe we would assign this to a QA agent or to maybe a dedicated security agent and just say every single day you should be running a check on the code and checking the integrations to make sure that we’re not leaving any secrets out there or we’re exposed to any, you know, cybersecurity attacks or anything like that. So now it says, “After creation, Paperclip takes you straight to the
16:49 — trigger setup for schedules, webhooks, or terminal runs.” So if I go ahead and create this, okay, well I actually can’t assign it to the QA agent because I haven’t yet approved that. You know, you can see it’s in my inbox. So let’s just for now go to the CEO, it should be able to delegate this to the right agent, but I’m going to go ahead and create the routine. But now this is super cool because we can make it active or inactive, we can set the schedule right here, and we can see all of the runs and all of the activity. So I think this is just super super cool. And here you go, you can see that we have our hire QA agent message right here, and we can go ahead and just approve this. Now that is
17:20 — how it’s set up on default, but you can change this if you want. So if I go down here to my settings, you can see require board approval for new hires. And this is where you can turn that off and just give your CEO the ability to just make hires and continuously grow the business. So let me show you guys something that I think is really really cool and there’s a lot of potential with this. So we have our organization, right? Now think about in a real business scenario, sometimes companies will acquire other companies because they have a really good team and they have like good frameworks and stuff. And that’s essentially what you can do is you can import companies, as you can see
17:50 — right here. So if we go over to the repo, you can see that if we go to the Paperclip company section, this is pre-built company templates that you can import into Paperclip. So if I click on this, you can see now as we scroll down, we have a growing catalog of ready-to-deploy agent companies or, you know, agent orchestrations that we can just put in. So we have G stack, we have superpowers, we have agency agents. We have all of these different things that we can actually just import, and it comes with agents, skills, knowledge. So for example, with G stack, it comes with a CEO, a CTO, a QA engineer, release
18:22 — engineer, staff engineer, and all of these skills are already in there. And there’s so many different ones that you can browse, whether you want to have, you know, an AI agency, whether you want to have, you know, some sort of autonomous AI intelligence company. You know, this one is specialized for scientific research, and there’s all of these different types of, you know, employees with specific knowledge. You can see these are named after actual doctors. So anyways, I just wanted to show you guys that if you wanted to get in there and play around with it, it’s definitely a cool way to start from, you know, like 50 rather than just starting from zero and building out your own team, you can just bring in a team. Like
18:52 — this one has literally 48 agents, which is crazy. So at this point, we still have these agents running for us. We still have, you know, the QA agent just finished something up. I could check the inbox here and see what’s going on. And if you wanted to have them keep working on stuff, once again, you’d make a new issue, or you could just comment and they would wake up, read your comment, and keep going. So if they have anything where they have like a specific blocker or they need something from you, they will ask for it. Now the final thing I wanted to talk about was secrets or API keys. And what you’ll notice is there’s not really a spot in here to have like
19:23 — an environment variable section or like a secret key manager, even if you are in your overall settings rather than just your project level settings. Now when I was talking to Cloud Code about this, it understood that there actually is something inside of Paperclip to hold environment variables. And that’s why it’s really nice to have this because it’s able to help me do that. So it’s able to create the secrets, it’s able to tell each agent where they live and how to use them, and that’s how my agent can use things like potato or, you know, nano banana two. And just to end off here, I’m showing you that I can say
19:53 — what companies do I have in Paperclip and what do they do, and it’s going to list them out. We have AIS Content and we have Proof Shot. We can see all the tasks, we can see things, and it can look at everything. So, I hope you guys see the value in being able to use Cloud Code to help you get all of your, you know, actual companies inside of Paperclip set up and optimized. So, anyways, that is going to do it for this one. Just a reminder that if you want that full sort of like road map resource guide on how exactly you should be setting this stuff up, if it’s your first time, then join my free school community. The link for that’s in the description, and you can get that for
20:23 — completely free. But, if you enjoyed the video or you learned something new, please give it a like. It helps me out a ton. And as always, I appreciate you guys making it to the end of the video. I’ll see you on the next one. Thanks, everyone.