The argument in one line.
Claude's new Code Remote Control makes OpenClaw obsolete by letting non-developers build a phone-controlled AI agent in under 10 minutes using a simple folder structure and onboarding prompt.
Read if. Skip if.
READ IF YOU ARE…
- A non-developer who tried OpenClaw, hit setup friction or bugs, and wants a phone-controlled AI agent without wrestling infrastructure.
- Someone running a small business or solo operation who needs a personal AI co-worker to handle context-aware tasks from mobile, not desktop-only.
- A creator or operator with existing Claude workflows who wants to extend those into a controllable agent without learning new platforms or languages.
SKIP IF…
- You need enterprise-grade agent orchestration or multi-agent workflows — this is single-agent personal use, not production scaling.
- You're already comfortable with OpenClaw's setup and running it daily — this video is about migration convenience, not new capabilities.
- You work primarily in environments where Claude Code or the Claude app isn't available (air-gapped networks, restricted tooling).
The full version, fast.
Anthropic's new Claude Code Remote Control feature replaces OpenClaw by letting you continue a local Claude Code session from your phone, tablet, or browser without moving anything to the cloud. The method is to treat your agent as a folder: a CLAUDE.md for behavior rules, a context folder holding user.md, business.md, and ICP.md profiles that load every session, a memory log for continuity, a review folder where finished work lands for human approval, and a .claude skills folder containing a voice profile. A single 400-line onboarding prompt interviews you, ingests any existing profile documents, and scaffolds the system in under ten minutes. Run the terminal continuously, isolate risky agents on a dedicated machine, and schedule recurring tasks through the review folder until native scheduling arrives.
Free for members
Chat with this breakdown — free.
Sign in and you get 23 free chat messages on us — ask for the hook, quote a framework, find the exact transcript moment, generate a markdown action plan. Bring your own key when you want unlimited.
Create a free account →Where the time goes.
00:00 – 01:22
01 · OpenClaw is dead
Alex acknowledges OpenClaw was groundbreaking but says Anthropic was always going to supersede it. He predicted it from day one.
01:23 – 02:50
02 · What Remote Control means
Strips developer language: text your agent from your phone while it runs locally. Nothing goes to the cloud.
02:51 – 07:22
03 · The folder system
CLAUDE.md (behavioral rules), context folder (USER.md, BUSINESS.md, ICP.md), memory logs, review folder, skills folder with voice profile.
07:23 – 13:17
04 · Setup wizard demo
Pastes the 400-line onboarding prompt into a fresh Claude Code session. Wizard interviews him; he drops existing md files into profiles instead of answering manually.
13:18 – 15:34
05 · Testing the agent
Opens alex-agent folder in Claude Code desktop app. Says hey there. Agent reads all context profiles and responds.
15:35 – 17:19
06 · Activating Remote Control
Types /remote-control in the terminal. Shows the Claude mobile app on iPhone connecting to the live session.
17:20 – 18:44
07 · Live phone demo
Texts the agent from phone; it responds in terminal. Both surfaces synced in real time.
18:45 – 22:22
08 · Limitations and closing
Terminal must stay running 24/7. No native scheduling; review folder is the workaround. Recommends dedicated Mac mini. Closes with Substack CTA.
Lines worth screenshotting.
- Claude Code Remote Control kills the main feature differentiation of OpenClaw — phone control of your agent — without requiring a separate Mac Mini setup.
- A context system built from USER.md, BUSINESS.md, and ICP.md loads agent knowledge at every session, making every conversation aware of who you are and who you serve.
- A review folder where agents deposit finished work gives you an asynchronous queue to process daily without needing proactive Telegram push notifications.
- OpenClaw's core contribution was proving that non-developers could command AI agents through their phone — Remote Control inherited that insight built natively.
- A 400-line onboarding prompt that acts as a wizard — interviewing you and then writing all the context files — collapses hours of setup into a single session.
- Long-term memory logs written by the agent to its own session folder give the system continuity across days without a separate vector database.
- The difference between OpenClaw and Remote Control is not capability but distribution friction — Remote Control requires no additional hardware or app.
- Naming context files USER.md, BUSINESS.md, and ICP.md makes the agent's knowledge categories explicit and human-readable.
- Borrowing the best architectural ideas from open-source tools (OpenClaw's memory and context structure) and layering your own methodology on top is faster than building from scratch.
- A local Claude Code agent controlled from your phone still runs compute on your desktop — all the benefits of mobility without cloud execution costs.
- Non-developer Claude Code setups benefit from named folders (context, review, memory) that mirror the agent's mental model rather than arbitrary project directories.
- Setting up Remote Control takes under ten minutes from a clean Claude Code session when you start with a well-structured onboarding prompt.
The non-developer AI agent template.
What to steal
Alex built a complete, phone-accessible Claude Code agent system that any non-developer can copy in 10 minutes -- and the mental model behind it is exactly where JoeFlow is heading.
- The USER.md / BUSINESS.md / ICP.md trinity is a ready-made CLAUDE.md template for JoeFlow Session workspaces.
- The review folder workflow is what JoeFlow batch queue does automatically -- use this as the explanation pattern for non-technical users.
- Think in folders is the mental model unlock: AI systems are just markdown files portable across platforms. Build this into JoeFlow onboarding.
- The 400-line setup wizard prompt is a smart lead magnet format -- a single paste that builds a full system and sells your methodology.
- His audience framing (non-developers who see Claude Code potential but feel locked out) is exactly JoeFlow ICP. Study his language for copy.
Terms worth knowing.
- Claude Code
- Anthropic's command-line coding agent that runs in a terminal, reads project files, and executes tasks against a local working directory. Originally aimed at developers but increasingly used for general knowledge-work automation.
- Remote Control (Claude Code)
- A feature that lets you continue a local Claude Code session from a phone, tablet, or browser. The agent keeps running on the local machine while messages are relayed through the Claude app.
- OpenClaw
- A community-built open-source wrapper that gave Claude a persistent personal-agent setup with Telegram messaging and remote control. Popular before Anthropic shipped native remote control.
- Coworking agent
- A persistent AI assistant configured with deep context about a single user, their business, and their audience so it can act as an always-on collaborator across daily work, rather than a stateless chatbot.
- CLAUDE.md
- A markdown file Claude Code automatically loads at the start of every session. It holds behavioral rules and project-wide instructions that shape how the agent operates inside that folder.
- Context profile
- A markdown file describing one slice of background the agent should always know, such as the user, the business, or the audience. Loaded into every session so the agent never starts cold.
- USER.md
- A context file detailing who the operator is — name, location, role, expertise, communication style — so the agent can address them and make decisions in alignment with their preferences.
- BUSINESS.md
- A context file describing the operator's company, products, pricing, clients, and offers. Gives the agent the commercial backdrop it needs to write copy, plan work, or handle decisions on-brand.
- ICP (Ideal Client Profile)
- A written description of the specific audience or buyer being targeted, including their pains, objections, and language. Used so the agent can write and reason in terms that resonate with that group.
- Memory log
- A running file where the agent records notes from each session so future sessions can pick up with continuity. Mirrors the short-term and long-term memory patterns popularized by OpenClaw.
- Skill (Claude)
- A reusable capability packaged in a specific folder format that Claude can load on demand. Lives in a .claude folder and can be reused across projects.
- Voice profile / Voice DNA
- A skill that captures a person's or brand's writing tone, rhythm, and vocabulary so the agent can produce copy that sounds authentically like them. Critical for ghostwriting and creator marketing.
- Markdown (.md) file
- A plain-text file format that uses simple symbols for headings, lists, and emphasis. Used here because AI agents read it cleanly and humans can edit it in any text editor.
- Terminal
- A text-based interface for issuing commands to a computer. Claude Code runs inside it, and the remote control session is started from a terminal window that must stay open.
- Slash command
- A shortcut typed into a chat or terminal that starts with a forward slash to trigger a built-in action, such as /remote-control to enable phone access to a session.
- WhisperFlow
- A voice-dictation app that turns spoken words into text inside any application. Useful for dumping long, unstructured context into an AI agent without typing it all out.
- Onboarding prompt / setup wizard
- A long prompt designed to make an AI agent interview the user step by step and build a configured workspace from the answers. Replaces manual file-by-file setup.
- Mac mini
- A small desktop computer from Apple often used as a dedicated always-on machine. Here, an isolated one runs the agent so it cannot touch the operator's primary computer or accounts.
- Scheduled jobs
- Automated tasks that run at set times without manual triggering, such as a daily morning report. A feature OpenClaw supported that native Claude Code remote control does not yet offer.
- Antigravity (Google)
- Google's coding-agent product, positioned as a competitor to Claude Code. Mentioned as another platform a context folder could be ported into.
Things they pointed at.
09:11toolWhisper Flow
Lines you could clip.
00:00
“Anthropic is about to kill OpenClaw with their new remote control release for Claude Code.”
02:22
“This literally means that you can speak to your agent through your phone.”
07:22
“Way easier than OpenClaw. Like, forget about OpenClaw.”
13:54
“Start looking at them as folders. Folders of markdown files, folders of skills folders.”
Word for word.
Read-along
Don't just watch it. Burn it in.
See every word as it's spoken — crank it to 2× and still catch all of it. The same dual-channel trick behind Amazon's Kindle + Audible.
metaphoranalogy
00:00Anthropic is about to kill OpenClaw with their new remote control release for Claude code. And in today's video, I'm gonna show you how you can start to take advantage of this and build your own coworking agent right now. So if you've been waiting, if you were, you know, deterred from trying out OpenClaw or setting up your personal agent, now is the time to do it because we could do it a lot easier with Claude code.
00:24If you've been watching this channel for the past few weeks, you know I was absolutely on the OpenClaw train, And I think what OpenClaw did and what it is doing, uh, was incredibly important and groundbreaking and deserves a lot of credit for getting us to this point where we're at today.
00:40But since day one, I knew, and I'm sure just like a lot of you know, especially anybody who's involved in the AI industry, we all knew that this OpenCloud thing was only gonna last for a period of time until one of the big players likely was always gonna be anthropic. Of course, we're we're gonna come in and just release their own release that that took care of this.
00:59And that's because a lot of what OpenCloud does is just built on top of, uh, Claude code or or, um, Anthropix models. So, you know, it was just a matter of time before that happened, and it happened sooner rather than later. Right?
01:12Before we get into this, ignore the crazy mess behind me. I just moved to a new apartment, so it's still a crazy mess. But I wanted to get in here and do this because it is, uh, super exciting.
01:21Alright? When it comes to OpenClaw, you know, I built it.
01:25I loved it. I was using it every day. Still still am right now as I'm experimenting with different ones, but it was a nightmare for many people to set up.
01:33There are a ton of bugs. It got annoying. A lot of the times, there were really annoying bugs.
01:38So, uh, I was super happy to try this out right away, and it's been, um, you know, a really good experience so far with the new Clogcode remote control. And I built my own system inside of it with my own methodology, so that's what we're gonna do today. You're gonna be building this AI coworking agent.
01:54Alright? So if we look here at the screen, we can see the ClawdCode docs for this new remote control, which allows you to basically continue a local ClawdCode session from your phone, tablet, or any browser using remote control.
02:07So like Anthropic always does, there's a lot of developer heavy language inside of their stuff. So you have to you know, for me and my audience of nondevelopers, you have to get past that and think about what this actually means for, like, you know, normal people who are not or nondevelopers.
02:22Right? And when it comes to this, this literally means that you can speak to your agent through your phone. That's basically you could text your agent through your phone.
02:29You can control your agent through your phone, and it's super easy to do this because you can do it right inside of the Claude app now. Alright? You see when you start a remote control session on your machine, Claude keeps running locally the entire time, so nothing moves to the cloud.
02:43You can work from both, you know, devices or surfaces at once, um, and it it works really great. Alright?
02:51Now there's no one way to set this up when it comes to nondevelopment work, alright, when it comes to setting up your CLOD code folder for nondevelopment work. So what I'm gonna show you and give you is how I do it in my day to day work, my methodology that I have been developing for over a year now that always gets plugged into these new tools, and that is how mine runs.
03:09So here, if you see in my folders here, I have a Claude agent. This is what it's gonna look like when we get done building it out. This is the folder that we're gonna be operating inside of with Claude code.
03:20You could see alright. There's a couple of things going on here. Right?
03:23We have the Claude dot m d file, which, you know, you are definitely familiar by now. And this is our behavioral instructions.
03:31Alright? So this is you know, you're gonna have all of this written for you when you set it up.
03:36I'm gonna give you a complete, uh, you know, prompt that you can paste in the Claude code, and it's going to shoot up like a onboarding wizard that you can build out your entire system. And it's gonna write this MD file for you, but it contains all the behavioral rules inside. Alright?
03:51And it tells your system exactly how to run and what to look at. Besides that, the next most important folder we have is the, uh, context folder. Right?
04:00And I borrowed a lot of these ideas off of OpenClaw. That's what's so great about all these different tools and all these different, you know, open source methods being released is that we can build on top of them. Right?
04:09So I combine my, you know, own methodology, my own techniques with what I could, you know, use from OpenClock.
04:15And what we got here now are three main context profiles that I have built into my system that your agent is going to read and load up every time. The first is the user dot m d, which if you're familiar with OpenClot, you know you know that is a major part of that.
04:32This tells your agent everything about you, uh, who you are, what you do, etcetera. Besides that, we have a business .md context profile. This one's super important.
04:43This is going to detail your business, you know, everything about your business, whatever it is you do, whatever it is you offer, pricing, etcetera. For me, this is, like, detailing all about my LLC, about my content business, about my clients, ghostwriting, etcetera, etcetera, so that my agent always knows what our business environment looks like.
05:02And then the third most important one is your ICP or your ideal client profile. Other words, your audience, who it is that you're targeting. So, you know, this will depend on what your role is or what your work is for me.
05:12For example, in my content businesses, I'm targeting nondevelopers who use AI for work, especially digital writing and stuff like this. So that is all detailed in this ICP as well. Alright?
05:23Now all three of those get loaded into your agent at every session so that it knows all your context. Besides that, we also borrowed from, uh, OpenClot the memory logs so that your agent keeps logging its own memory, and this gets pushed into its session each time, and then a long term memory as well. Alright?
05:42So we're gonna set this entire system up, and you're gonna be able to do this in, like, ten minutes. Now besides your context folder, which provides all the context to your agent, we have a couple more things here. We have a review folder, which, of course, this is optional, but I built this in because here's the deal.
05:58Claude is are the Claude code remote controls. If you're familiar with OpenClaw, it's not going to be proactively texting you like OpenClaw would through Telegram or wherever.
06:10So instead, workaround for this kind of is to create a review folder that every time I'm working with my agent and it finishes something, you know, whether that be a newsletter or a YouTube script or thumbnails or whatever, it is putting all of this inside of a review folder. It's organizing it all, I can open that every day and kinda just work down and review what I want and publish, etcetera.
06:30So we have a review folder, and our agents gonna automatically know to put this all of its work inside of a review folder instead of placing it, you know, in final locations. I like to throw everything into a review folder. This keeps everything organized, and then leave it up to me, the human, to actually move it to its final home.
06:45Okay? And then lastly, we have, of course, our dot clod folder with our skills inside. And the only skill I have in this one is my voice profile, which I've talked about a lot of times before.
06:57This is one of the most crucial skills you can have for any type of, you know, marketing work, creative work, newsletter writing, digital writing, anything like that. You need to have ghost writing, a voice profile so that Claude can write in your voice or a client's voice, company's voice, etcetera. And then we'll we'll put a bunch of more skills in there over time.
07:14Right? But this is what the entire system looks like here. Okay?
07:18Now to do this, like, to set this up yourself very easily, way easier than OpenCLaw. Right?
07:24Like, forget about OpenCLaw. This is so much easier. You're gonna need to have Claude code.
07:28Okay? So here on the top, you can see my terminal, and you are going to launch up Claude code.
07:35So we have Claude code here. Alright? And what you're gonna do is grab the prompt below that's in my substack.
07:42This is a big, like, 400 line, I think, setup prompt, and this is going to set up the entire onboarding of your system. So if we can see here I don't let me see if I can make this a bit bigger. You could see here we have, um, a setup prompt to set up a personal coworking agent.
08:00Alright? It's going to walk through a bunch of questions. It's gonna set up all of your different context profiles.
08:06So that business profile, that ICP profile, and your user profile, this onboarding through ClaudeCode is gonna set it all up for you.
08:15You're gonna name your agent. It's gonna build everything in a folder, and then you're gonna be ready to go. Alright?
08:19So I'm gonna take this prompt. You're gonna grab this below, and you're just gonna paste this into a fresh new quad code session. Okay?
08:27Now let me I can get rid of this now, make this bigger so we can see everything here. Alright.
08:35Now if you see once you put this prompt in here, you're gonna get, hey. Welcome to the coworking agent setup. I'm gonna walk you through setting up your personal Claude code agent in AI that actually knows who you are, what you do, and who you do it for.
08:50By the end of this, you'll have a system that remembers your context every time you open Claude code, and then it's gonna ask you questions. It's gonna ask you questions about yourself, about your business, about your audience. Those are what are gonna set up your context profiles.
09:02Then it's gonna ask about your writing voice or if you have a voice DNA skill already, and then it's gonna name your agent and build everything for you ten minutes. Alright? First question, it's gonna ask you is right here.
09:13It's gonna wanna get to know you, so you're gonna tell it about yourself, share whatever feels relevant, who you are, where you are, time zone, background, what you do for work, what your company is, expertise, how you like to work, how you communicate. These are just some things I threw in there just for ideas, but you can be as detailed as you want about anything that you want your personal agent to know about you.
09:31Alright? So, you know, uh, one recommendation I would have is to download something like Whisper Flow so you can use voice dictation here. Hold in your little dictation button and just start rambling for ten five, ten minutes, all about your business, all about everything.
09:45You know, just vomit all of that onto the chat and let Claude, your agent, do the work organizing that into a profile. You can also drop documents in there so you have documents about your yourself or biographies, your LinkedIn page, social media, anything like that. You can drop all that in as well.
10:01And lastly, if you're like me who already has a bunch of these profiles, for example, I had the user dot m d. I have the, uh, business profile.
10:10You can just drop that as well. Alright? So that's what I'm gonna do.
10:13So I'm gonna grab all my profiles I already have here in the downloads folder just because this will also save a bunch of time, but you can go through all of this or you can just drop them if you already have it. So I have a user dot m d already.
10:24I'm just gonna drop it, give it to Claude so it's gonna be able to read all of this and understand who I am.
10:32It's gonna read the file. Alright. And now it it finished up here and now it knows about me.
10:38It knows who I am. Alex Fron twenty eight based in Sao Paulo. Gotta fix that.
10:42I'm not in Sao Paulo, but I am in Brazil. Solo operator, expert in AI cowriting systems, ships fast, blah blah blah blah blah.
10:49So all sorts of stuff about me it knows now. Okay? Now it's gonna ask about the business.
10:53Tell me about your business. This helps me understand what you're building. Alright?
10:57Business, I'm just gonna drag my business MD file in here that I already have. This is all about my business. This about my LLC structure.
11:04This is about my content business. This is about my service based businesses, uh, ghost writing, all of that.
11:09It's gonna get all of that context here. Alright? Alright.
11:13So now it has this. It knows my LLC. It knows what I do.
11:17It knows my core product is the co writer system, um, my funnel, differentiators, plans I have for the future.
11:23And then lastly is the audience or the ICP. So you're going to, you know, work through that interview with Claude or just drop any documents you have here, and this will close out our context profiles here. Alright.
11:36So you can see it just finished. We've got all three profiles locked and loaded, ready to go. Here's a quick summary of me.
11:43You know, it knows my ICP is nondeveloper professionals. Core frustration is seeing Cloud Code's potential, but feels locked out because they're not developers. The core shift, primary objections, and some language tips.
11:54And then lastly, it's gonna ask for your voice. So it's gonna ask you if you have an existing Claude skill for writing your voice, voice DNA or voice profile, whatever you wanna call it. And if you don't, it's okay too because you can just, you know, skip this part.
12:08You can tell you don't have it. And as you communicate with your personal agent, it will learn your communication preferences. Okay?
12:15It will learn how you talk, and it'll work on that itself. But I already have a skill, so I'm just gonna drag it in here as well. Voice profile.
12:23Done. It will recognize this. It needs to be in the format of a skill, which it's already built out for Claude.
12:29So it will just be able to basically just duplicate that right into our new system. And a quick note on this, you can have multiple voice profiles. Right?
12:37Maybe you have one for yourself, for various clients. You can explain all of that. The system is completely flexible just like any of other ones, OpenClock, etcetera.
12:44You can completely customize this. Alright? So it has the full voice profile.
12:49And lastly, it's gonna ask for a name. Choose whatever. I'm just gonna put Alex agent for right now, and that will name the folder that.
12:57So the folder is gonna be named to that. The agent is gonna take on that identity. And then going forward, you're always gonna be working in this folder.
13:03Alright? Gonna give it permission to now create the full directory structure for this, uh, system, for this agent.
13:13Alright. So it took just a couple minutes, and it finished less than ten minutes easily. And my coworking agent is now set up.
13:21Here's what Claude created. We have an Alex agent folder with everything inside of it. Let's take a look here.
13:26When you stop, you could see the Alex agent folder now with the entire system built inside of it. We have the Claude dot m d file. We have the, uh, context profiles, the memory, and the review folder with a little welcome in here.
13:41Okay? So that is now all built out, and we can actually test it out. So now that this folder is built, you know, this is one way you have to look at building these systems, uh, with AI.
13:54You need to start looking at them as folders. , folders of skills folders. Because if you start looking at it that way, then you realize it could be plugged into different platforms.
14:03So this same folder, we can use it in regular Clawd code on the app. We can make a few adjustments and use it inside of Cowork. We can now use it inside of the remote control app or remote control.
14:15And then if you really wanna go out there and say you just really hate Claude, you can also adjust this to work for ant Google anti gravity, you know, etcetera. So, uh, now let's test this out though.
14:25The first way I wanna show you is just is just that. We're gonna go into the Claude app and just test out the system, you know, by itself.
14:33We have not even we've not even gotten into the remote control yet, which makes it crazy good. But alright. So we're inside of Claude code now, the desktop app, and we have this Alex agent folder opened.
14:45We're just gonna say, hey there. And you're gonna see it is going to read all the context profiles and know everything about me right away. Alright?
14:53And that's how this works. Every session you can see is gonna load up the context. So it's reading my user dot m d, business dot m d, icp.
15:01Uh, it's reading memory. It's reading, uh, the skills the descriptions of skills it has.
15:07Here you go. Context loaded. Looks like this is a fresh setup day.
15:10System just spun up with your context profiles and voice skill. And then we could work right here inside of Cloud Code. So if you just wanna work like this in this familiar interface, you absolutely can.
15:21I'm sure eventually. Of course, there's gonna be way more features coming here. You'll probably be able to remote control this as well.
15:27But if we do wanna use the remote control, we need to use the terminal. Alright? So what we're gonna do here is go back to the terminal, and we need to we're gonna terminate this one and start a brand new one.
15:39But this time, we wanna open it inside of our folder. Okay? So remember, every time that you're working inside of the folder, that is, you know, the the the agent.
15:48That is the system. So here we have Alex agent. I'm gonna open a terminal in this folder, okay, and launch it up.
15:55And now same situation here. Okay? We have Claude code, but now we have it inside, uh, the terminal inside of this folder.
16:05And same situation here, we can just say, hey there. Get it all booted up here, and then we will move on to the coolest part, which is controlling it from your phone.
16:13So it's gonna load up its context. Same situation, searching for all the files. Alright.
16:20So now you can see our agent running inside of Cloud Code, the our personal agent, and it's a completely fresh, uh, workspace. And to to now enable this remote control, we need to put slash remote control.
16:34Right? You see remote control connect this terminal for remote control sessions. You need to do that.
16:39It's gonna connect, and now we can open it from our phone once it's done. So you have a couple ways of doing this. You can follow this URL.
16:46Uh, you can scan a QR code, and you can also just open up your Claude app on your phone, which is what I'm gonna do here. So you can see my phone opened up here. See, can I make this bigger?
16:56No. I can't. This is my phone inside of the Claude the Claude app with the Claude code.
17:02And now if we refresh, we should be able to see the, uh, the new the new session.
17:09Alright. I had to close down the app and refresh, but now that it's refreshed, you could see we have this in progress, and then this is our current session.
17:18So if you click this here, this would be on my phone. Right?
17:21I could be away from my computer. I could be anywhere. Now you can see that we're looking at the same thing.
17:25So now we're remote controlling it. Same idea if you were using OpenCon stuff and you were using Telegram to talk to it.
17:32It's the same idea now. I could be at the beach. I could be wherever outside, and I could be talking to my agent doing work on my computer here.
17:38Right? So if you come here and I'm gonna text it on the phone, you're gonna see it work here in the terminal. So I'm gonna say, tell me about myself and your infrastructure, and we'll see what it can tell us about itself.
17:56Now you can see I typed it here on my phone, and if you look in the left inside of the terminal, I'm not touching it. It's doing it itself. Right?
18:03So same idea here. We have now bridged the gap between our, uh, device, our computer, and our phone. We could talk to our agent wherever we go.
18:12And here you could see it popped up on the terminal, and it's probably gonna take a second to pop up here on the phone. And there it is. You're Alex McFarlane, 28, living in Sao Paulo, Brazil, started writing, blah blah blah.
18:23You run your LLC with the co writer system, my infrastructure, user.md, business.md, etcetera, etcetera.
18:32That's to set up what do you want to build, and now you can start building. You can upload your skills to this folder, all of your Claude skills, your writing systems, your whatever work it is you do, and now you have a personal agent that you can take with you anywhere you go. Now before closing this out, a couple main main points, of course, that you should know.
18:50First, this terminal needs to be running twenty four seven for you to be talking to it through this remote control. You can shoot up new sessions and stuff, but it's good just to leave this terminal running and then talking to it on your phone, um, to avoid any problems. And at least that's from my understanding, this might be this might be different.
19:08This might be changed. You can comment below if it's easy to launch up a new session, um, from your phone. And then secondly is, I'm running this on my main computer right now, my day to day work computer.
19:20With OpenClaw, if you know, I literally have OpenClaw running on a completely isolated Mac mini. Okay?
19:27Um, you know, that that whole hype, I did it, and it wasn't hype. It works really well. I have an a Mac mini dedicated Apple IDs, um, emails, browser control, everything dedicated to one Mac mini that's clean, doesn't have all my, you know, information on there.
19:42And I'm gonna do the same thing now with this Claude code agent. I'm gonna put it on its own Mac mini so I don't have to worry about it messing with my day to day work, deleting anything, you know, it can run, it can control its browser, and I can just keep experimenting without any worries. With that said, it's not necessary.
19:58It's probably not as necessary as it was with OpenClaw because without a doubt, you know, there's still risks with this.
20:06You need to be careful. It has control of your your system. Anthropic itself says to be careful.
20:11But without a doubt, this is miles ahead safer than than OpenClaw, um, like, a lot safer than OpenClaw. Right?
20:17So that's another important point. Feel free to run this on its isolated device. I would always recommend that.
20:22But, also, this is safer to run, you know, regularly with some, of course, being smart about it and and careful. Thirdly is, as of right now, this can't schedule jobs.
20:34So, for example, on OpenClaw, I was having it text me every day and analyzing my YouTube analytics and giving me a report every day, right, at 09:00 in the morning, for example. Uh, Cloud Code agent right here can't do this right now.
20:48They did just release for co work a new a new feature where you can schedule things, but it's not here for this right now. But, you know, that's okay.
20:57I'm sure it's coming at some point. They're pushing this stuff forward every day. And that's why I have the review folder here as a workaround so that, for example, instead of it, um, messaging me every day, I can just, you know, have it review certain things or have it carry out certain tasks and drop it in the review, and I can review periodically.
21:17There's probably workarounds to set up jobs. If you know, uh, let me know. Um, but, yeah, that's that's pretty much it.
21:24Um, that's that's how you set up your own personal agent with Claude code now using the new remote control. Pretty pretty sure in my opinion, OpenClaud dead, done, dead in the water.
21:35ClaudeCode will take this over. So go build it out way easier if you're ever scared of OpenClaud, you know, now is your time to try it out.
21:43And as always as always, whenever you have all this hype about these these things coming out, for example, Open Claw, I get into the hype. I follow it because this is what I do.
21:53Everything I do revolves around AI. But I tell you all the time, don't freak out.
21:58Just wait for the stuff to mature because it's not gonna take long. And as you can see, this took, like, less than a month for Claude for Anthropy to come out with their own version of something like this and and working towards this. So all you have to do is wait for it to mature.
22:10You're not gonna miss out if you don't want to. And lastly, grab that setup prompt, subscribe to my Substack, subscribe to this YouTube. We'll be building a lot inside of here because it's just a lot better than OpenClaw in my opinion.
The Hook
The bait, then the rug-pull.
Alex McFarland opens with a declaration, not a question: Anthropic just made OpenClaw obsolete. Then he spends 22 minutes proving it live -- from setup prompt to phone-controlled agent.
Named ideas worth stealing.
02:51model
The Context Folder System
- USER.md
- BUSINESS.md
- ICP.md
Three markdown files loaded into every Claude Code session so the agent always knows who you are, what your business is, and who your audience is.
Steal forCLAUDE.md template for JoeFlow Session workspaces
05:57concept
The Review Folder Workflow
Instead of push notifications (not yet supported), all completed work drops into a review folder. Human reviews and moves to final destination on their own schedule.
Steal forExplaining JoeFlow batch queue to non-technical users -- same mental model, automated
13:54concept
Think In Folders
AI agent systems are just folders of markdown files. Portable across Claude Code, Cowork, remote control, or other providers.
Steal forJoeFlow onboarding and portability positioning
How they asked for the click.
VERBAL ASK
22:01link
“Grab that setup prompt, subscribe to my Substack, subscribe to this YouTube.”
Verbal CTA only, no on-screen card. Substack is the primary lead-gen destination -- setup prompt is the bait.
MENTIONED ON CAMERA
FROM THE DESCRIPTION
PRIMARY CTAWhere the creator wants you to go next.
OTHER LINKSAlso linked in the description.
Visual structure at a glance.
