A zero-employee founder rebuilds his LinkedIn-to-email funnel live, showing every tool he's wired straight into Claude Code.
Posted
6 days ago
Duration
Format
Interview
educational
Views
3.4K
53 likes
57 · 43
Big Idea
The argument in one line.
A solo founder can run a $250,000-a-month business with zero employees by treating Claude Code as an orchestrator wired into every tool, analytics, CRM, email, deploys, even health data, instead of hiring a team to run each one.
Who This Is For
Read if. Skip if.
READ IF YOU ARE…
A solo founder or small-team operator who wants a concrete map of which SaaS tools to connect to an AI coding agent and what each connection actually buys you.
A creator running a LinkedIn content-to-email funnel who wants to see the exact steps (infographic, landing page, tag, sequence) built live rather than described.
Someone deciding between keeping Claude Code running locally on a machine or moving to an always-on cloud agent, and wants to see both approaches side by side.
SKIP IF…
You're looking for a step-by-step technical tutorial with exact commands and code; this is a walkthrough narrated at a business level, not a build-along.
You have no existing tool stack (CRM, email platform, analytics) to connect; most of the value here comes from wiring up tools you already pay for.
TL;DR
The full version, fast.
James McAulay runs a $250,000-a-month, zero-employee business entirely through Claude Code, with every business tool (analytics, CRM, email, deploys, even health data) wired in as a connected source. The core mechanism is a reusable skill: Claude asks clarifying questions, builds a greyscale wireframe, checks live PostHog data before proposing changes, ships through Vercel with Sentry/Cubic/Playwright as safety checks, and wires new leads straight into Kit, Attio, and Apollo. A scheduled Telegram agent (built on Vercel's Eve framework) delivers a daily sales briefing and doubles as a personal health coach reading Garmin and TrainingPeaks data. The throughline is that a well-organized knowledge base and a handful of reusable skills turn Claude into an operating layer for the whole business, not just a coding tool.
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.
Cold open on the $250K/month, zero-employee claim, then a mind-map of every tool in the business feeding into Claude Code as the central orchestrator.
00:42 – 02:32
02 · LinkedIn infographics: HTML into an animated GIF
James's best-performing LinkedIn post is secretly an animated HTML page, captured frame by frame into a GIF, and he invokes his LinkedIn Infographic skill in the terminal to rebuild one live.
02:32 – 03:37
03 · WhisperFlow voice input and the 3-step lead magnet funnel
James dictates prompts by voice at roughly 150 words a minute and lays out the funnel he's about to rebuild: infographic, landing page, then a tagged signup in Kit.
03:37 – 05:34
04 · PostHog dashboards and clarifying questions before Claude builds
Claude reads live, free-tier PostHog dashboards to answer 'how is this performing?', then asks clarifying questions (story angle, real vs. generalized content, three-step vs. loop) before building anything.
05:34 – 07:53
05 · Wireframes first, then the LLM wiki second brain
A quick black-and-white wireframe pass catches layout problems early; James explains his business's history lives as two to three thousand organized markdown files Claude can search.
07:53 – 10:02
06 · SEO data and a Next.js codebase Claude can edit directly
Claude pulls live SEO and Lighthouse data to justify page changes, and because the site is a Next.js app in GitHub, 'create a new version of this page' is a request it can just execute.
10:02 – 11:09
07 · Vercel ships it, Sentry, Cubic, and Playwright check it
The deploy stack: Vercel for shipping, GitHub for version history, Sentry for error alerts, Cubic for AI code review before anything goes live, and Playwright so Claude can visually inspect its own work.
11:09 – 12:37
08 · PostHog data drives conversion tweaks, then a checklist templates the next page
Claude names specific conversion-rate tweaks using PostHog scroll and click data, then follows a checklist a previous agent run left behind to build the new lead magnet page.
12:37 – 14:28
09 · Kit.com: the email side of the funnel
James walks through the nurture sequence in Kit: the automatic delivery email leads with the free skill and mentions his paid program only once, near the bottom.
14:28 – 16:25
10 · The finished animated infographic and a live landing-page preview
The rebuilt infographic renders as a GIF; James critiques the first draft live (text too small, too text-heavy), and Claude previews the new landing page on localhost before deploying.
16:25 – 17:50
11 · Kit API automation and protecting deliverability with Resend
Claude holds a live Kit API key and creates the tag, the form, and the site wiring itself; transactional email routes through Resend to protect Kit's sender reputation.
17:50 – 19:03
12 · Attio and Apollo auto-enrich every new lead
The moment someone signs up, Attio creates a CRM record and Apollo enriches it with company size, estimated revenue, and title, so James can spot his ideal customer profile from an email alert.
19:03 – 20:41
13 · Jamie: a daily sales briefing on Telegram
A scheduled agent called Jamie messages James every morning on Telegram with his calls, pipeline state, and estimated revenue against target.
20:41 – 22:01
14 · Turning a folder into a cloud agent with Vercel Eve
James explains Vercel's Eve framework: one command turns a folder of instructions and code into an always-on cloud agent connectable to Slack, Telegram, WhatsApp, or email, and shows a before/after of tidying that folder for context.
22:01 – 25:37
15 · A health agent built from Garmin, TrainingPeaks, and Telegram
The same Eve pattern applied to personal data: a health agent reads Garmin sleep/HRV and TrainingPeaks workout data and answers questions like 'how was my run last night?' over Telegram.
25:37 – 26:44
16 · Structured CSVs make big datasets easier for Claude
Claude struggles with sprawling raw health data, so James consolidates weight, VO2 max, and training readiness into a single structured CSV, which also powers a Friday 'Weekly Retro' skill.
26:44 – 27:32
17 · Always-on cloud agents, and the OpenClaw/Hermes comparison
Andrew floats Eve as a potential 'OpenClaw killer' or 'Hermes killer,' and both agree the real barrier isn't capability, it's that almost nobody outside developers knows the setup exists.
Atomic Insights
Lines worth screenshotting.
A solo founder can run a $250,000-a-month business with zero employees by wiring every tool in the business into Claude Code instead of hiring a team.
LinkedIn gives animated GIFs the same reach as static images, and most people don't know you can post one at all.
An infographic built as an HTML page, then captured frame by frame, becomes an animated GIF cheap enough to make a new one for every post.
Configuring an agent to ask clarifying questions before generating anything catches the wrong assumption before it costs a redo.
A greyscale wireframe pass before the final design exists because layout, not visual polish, is the hardest part to get right, and it's cheap to iterate on early.
Eight months of business history can live as two to three thousand markdown files in an organized 'LLM wiki', giving an agent the same context a longtime employee would have.
Giving an agent live access to analytics, SEO, and performance data means a requested change comes with the actual metric behind it, not a guess.
Pairing an AI coding agent with automated review (a tool catching one or two issues on almost every deploy) is what makes it safe to ship AI-written code without a human gate on every change.
A documented checklist a previous agent run leaves behind turns repeat work into a template instead of a problem solved from scratch each time.
Routing transactional email through a separate service protects the deliverability of the marketing list that actually needs it.
Automatic lead enrichment at signup (company size, estimated revenue, job title) turns a bare form fill into a scored lead before a human ever looks at it.
A scheduled agent that reads a CRM, calendar, and to-do list can deliver a daily pipeline briefing over chat without anyone opening a dashboard.
An AI agent can choke on raw, sprawling personal data; consolidating it into one structured CSV is what makes it usable.
Turning a folder of instructions and code into an always-on cloud agent removes the need to keep a personal computer running in the background.
The biggest barrier to autonomous cloud agents catching on outside developers isn't capability, it's that almost nobody understands the setup or even knows the tool exists.
Takeaway
Wire every tool into Claude and it becomes the operating layer for the business.
WHAT TO LEARN
The value isn't any single automation, it's an organized knowledge base plus live API access to every tool, which lets one agent plan, build, ship, and follow up across the whole funnel without a human relaying context between systems.
02LinkedIn infographics: HTML into an animated GIF
An infographic built as an HTML page, then captured frame by frame, becomes an animated GIF cheap enough to make a new one for every post.
The workflow runs from a single reusable Claude Code skill: describe the infographic in a sentence, and Claude builds it from wireframe to finished GIF.
03WhisperFlow voice input and the 3-step lead magnet funnel
Dictating prompts at roughly 150 words a minute versus typing at 100 is enough of a speed gain that it changes how much a founder is willing to ask an agent to do.
A lead-gen funnel that fits in three steps (post a graphic, gate the deep-dive, tag the subscriber) is simple enough to rebuild in one sitting.
04PostHog dashboards and clarifying questions before Claude builds
Free-tier analytics dashboards, queried live, let an agent answer 'how is this performing?' with an actual number instead of a guess.
Configuring an agent to ask clarifying questions before building anything catches the wrong assumption before it costs a redo.
05Wireframes first, then the LLM wiki second brain
A quick black-and-white wireframe pass exists because layout, not visual polish, is the hardest part to get right, and it's cheap to iterate on before the real design is generated.
Eight months of business history can live as thousands of organized markdown files, so an agent can look up anything historical as plain text instead of it living in someone's head.
06SEO data and a Next.js codebase Claude can edit directly
Giving an agent live access to SEO and performance data means a requested change comes with the actual metric behind it, not a guess.
Because the whole site lives in one connected codebase, 'create a new version of this page' is a request an agent can just execute, not a spec someone hands to a developer.
07Vercel ships it, Sentry, Cubic, and Playwright check it
The deploy stack pairs an agent that writes code with independent checks: error tracking after deploy, automated code review before it ships, and a tool that lets the agent visually inspect its own work.
An automated code-review tool catching one or two issues on almost every change is what makes it safe to ship agent-written code without a human review gate every time.
08PostHog data drives conversion tweaks, then a checklist templates the next page
An agent can read scroll and click data to name a specific conversion tweak, instead of giving general design advice.
A checklist a previous agent run leaves behind turns repeat work into a template instead of a problem solved from scratch each time.
09Kit.com: the email side of the funnel
The strongest nurture email delivers the promised free value first and mentions the paid offer once, near the bottom, instead of leading with a pitch.
A single email to a list converting a few hundred people into lead-magnet downloads, which then feed workshop signups, is the actual mechanic behind a funnel that 'refills itself.'
10The finished animated infographic and a live landing-page preview
Reviewing an agent-built asset out loud in real time (text too small, spacing feels forced) is what actually gets it postable, not a simple pass/fail judgment.
Previewing a new page locally before deploying catches problems while they're still free to fix.
11Kit API automation and protecting deliverability with Resend
An agent holding a live API key for the email platform can create the tag, the form, and the site wiring itself, leaving only the nurture-email copy for a human to write.
Routing transactional email through a separate service protects the deliverability of the marketing list that actually needs it.
12Attio and Apollo auto-enrich every new lead
Automatic lead enrichment at signup can return company size, estimated revenue, and title, turning a bare form fill into a scored lead before a human ever looks at it.
A scored lead (title, company size, revenue) is what makes manual outreach worth sending instead of a blind follow-up.
13Jamie: a daily sales briefing on Telegram
A scheduled agent that reads a CRM, calendar, and to-do list can deliver a daily pipeline briefing over chat without anyone opening a dashboard.
Estimated revenue calculated from pipeline stage and probability, delivered daily, turns 'am I on target' into a number instead of a feeling.
14Turning a folder into a cloud agent with Vercel Eve
One command in a folder can turn that folder's context files and code into an always-on cloud agent, removing the need to keep a personal computer running in the background.
An agent finds context faster and burns fewer tokens the more organized its folder is, so cleaning up duplicate and stale files is a direct cost lever, not just tidiness.
15A health agent built from Garmin, TrainingPeaks, and Telegram
A personal health agent reading sleep and stress data can flag when someone is sleep-deprived and discount what they say accordingly, acting like a coach that understands physiology, not just the words typed.
The same agent framework used for a business, a folder, scheduled tasks, a chat channel, works identically for a personal health agent; it's the same pattern applied to a different data source.
16Structured CSVs make big datasets easier for Claude
An agent can choke on large amounts of raw, sprawling data; consolidating it into one structured CSV is what makes it usable, not a bigger model.
17Always-on cloud agents, and the OpenClaw/Hermes comparison
The real barrier to autonomous cloud agents catching on outside developers isn't capability, it's that almost nobody understands the setup exists or what it does.
Glossary
Terms worth knowing.
LLM wiki
A folder of thousands of organized markdown files that functions as a business's second brain, letting an AI agent search institutional history as plain text.
Skill (Claude Code)
A reusable, documented workflow saved so Claude Code can run the same multi-step process again without re-explaining it each time.
MCP (Model Context Protocol)
The standard that lets Claude Code connect directly to a tool's live data (like a CRM or analytics platform) instead of working from a static export.
Vercel Eve
A framework from Vercel that turns a folder of instructions and code into an agent that runs continuously in the cloud, connectable to channels like Slack, Telegram, or email.
Lead magnet
Free content (like an infographic or guide) offered in exchange for an email address, used to start a funnel.
ICP
Ideal Customer Profile: the specific type of lead (by company size, revenue, or role) a business is trying to attract.
“This founder is doing $250,000 a month in revenue with zero employees. Everything is done using Claude.”
cold open stat, no setup needed→ TikTok hook↗ Tweet quote
06:25
“The thing that I found is the hardest is getting the layout correct. Like sometimes Claude wastes a bit of space.”
honest, specific limitation from a heavy daily user→ newsletter pull-quote↗ Tweet quote
18:15
“When I see that email, I think, okay, this is my ICP. They have the money to spend on AI training.”
concrete lead-scoring logic in one line→ IG reel cold open↗ Tweet quote
22:10
“You're telling me you feel stressed, but you also averaged like six hours sleep a night. No wonder you're stressed.”
vivid example of an agent using context beyond the literal question→ TikTok hook↗ Tweet quote
26:51
“I think this could be like the kind of open the claw killer or the Hermes killer.”
bold, quotable competitive claim to close on→ IG reel cold open↗ Tweet quote
The Script
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.
17px
This founder is doing $250 ,000 a month in revenue with zero employees. Everything is done using Claude. He's going to show us his setup.
Presented by Zapier, the AI automation company. All right, what are we looking at here, James? So I thought I'd map out every single tool in my business and it shows that everything is actually connected to Claude Code.
I run my entire business now in Cloud Code, usually Fable or Opus. And when Cloud is down, I actually use GLM 5 .2 as like a Chinese cheaper model. And this enables me to do some pretty cool stuff.
For example, I've got every single tool in my content stack connected to Cloud. I've got all my analytics connected to Cloud. And I thought I'd show you how I create infographics like this.
So this is one of my best performing posts on LinkedIn in terms of qualified leads for my business. This is a classic, you know, comment the word compound and I will DM you a skill that improves your cloud setup every Friday. So I've had 400 comments on this post.
This entire infographic was made by Claude and this leads people into a funnel that then converts them to my business over time. Okay, I really like how this infographic has a pulse to it and some energy. Okay, you're in the business of teaching people how to use AI.
So let's see how you use it. What do you use to create this infographic? All right, so under the hood, this image is an HTML page.
This is almost like a custom website, animated, and then frame by frame turned into an animated GIF. I don't think many people know this, but you can share GIFs on LinkedIn and they still get the sort of reach and engagement of an image. So images right now are some of the best performing formats on LinkedIn.
So I have a skill that will basically, I will talk to Claude about the infographic I want. It will create a wireframe so I can say like, this looks good, change this. And then when I'm good with it, it will sort of flesh it out, animate it and then give me the GIF.
So shall I show you how I do it? Yeah, let's take a look. Let's do it.
Okay, so I run my entire business in Cloud Code in the terminal. You can do this in Cloud Codes and the desktop app. You could even do this in Cloud Code Work, but I think this is the fastest way to do it.
So there is a skill that I have created called LinkedIn Infographic. Oh, didn't mean to hit that. And this is going to take me through the process I just described to you.
I use Whisperflow. so that I can speak at something like 150 words a minute instead of typing at about 100. And we'll give people this and all the skills below in the description.
All right, go ahead. Everything. Yeah, everything that we're going to show today will be available afterwards.
All right. So I want to create an infographic that visualizes the LinkedIn infographic through to landing page and then through to mailing list. funnel that we have created.
So there are three steps here. There is step one, creating the infographic. Step two, creating the landing page, the lead magnet landing page.
And then step three is someone joining that page and being tagged correctly in kit .com. Ask me any questions you need to ask me and then let's create the wireframe.
I'm going to have that run. That's going to take probably like 30 seconds. And while I do that, I just want to show you like something else that I can do by connecting everything to Claude.
So if we just zoom this in and then I will show this here. I can ask Claude, you know, how is this performing? So Claude has created all of my dashboards for me.
I used to do growth at tech startups and in my own company. And it would take ages to actually make sure that like all the... insights were set up correctly and now i just say to claude like whenever you're building anything make sure everything's tagged and then go and build these nice dashboards for me so yeah all of this is on the free plan highly yeah i i you know you can see that i've hit like a limit for session replay but apart from that i'm still paying them zero dollars a month which is incredible to me like i big big post hog fan over here highly recommend um all right so My cloud is set up to always ask me for clarifying information rather than trying to guess.
So firstly, what is the story angle? What are they trying to take away? Do we want an agent built this funnel?
Do we want copy this funnel? Or do we want both? I think copy this funnel.
I think making this something that someone can take away themselves is valuable. Okay, do we want a three -step flow or a loop?
I think we want a vertical funnel and then statically animated, obviously animated. How real should the content be?
Let's do generalize so that someone can like apply this to their business. So it's going to probably create a wireframe in a second. Meanwhile, Claude here, because I'm using Fable, it's taking, you know, like a couple of minutes, but it is scanning every single landing page.
to see like conversion numbers for the original lead magnet. Okay, gotcha. This is like speaking to a growth analyst, right?
I've just said, how is it performing? So now we know that our benchmark for this new page is like, I'd say 45%. This has to convert at least 45 % to be as good as the other ones.
Okay, so Claude is building the wireframe for this infographic now. And that's part of your skill to not show the finished product first to do the wireframe first. Okay.
Absolutely. Yeah. Because it's really quick and easy to generate a sort of black and white wireframe.
And also the thing that I found is the hardest is getting the layout correct. Like sometimes Claude wastes a bit of space. Sometimes things feel too cramped.
Like I want to scan this wireframe and check that someone scrolling through LinkedIn can immediately understand it. We can actually, if we want to see the work in progress, we've got this. So three -step lead magnet funnel, post a graphic and then gate the deep dive and then tag everyone at the door.
I do like that it's actually building an infographic based on what it knows about your business. How does it know all this stuff about your business? Where is it saving the data?
All right. So this business actually, I started out consulting and I had everything in my own personal, GitHub repo on my personal website and then loads of people were asking me for help with agents so I created like an agent workshop and then I created a course and basically over time everything has like grown in this sort of second brain in my main folder so every single every single thing I've done in this business over the last eight months is documented somewhere in this second brain in like I think I've got two or three thousand markdown files now, all like neatly organized as a sort of, it's called an LLM wiki.
So anything historical, Claude can read as like a text file. Any live data, it's just, again, it's like plugged into every single tool. So it can just go and like kind of look up any live data about, you know, how are we doing things in kit?
In terms of analytics, like how is that page performing? Yeah, everything is plugged in so Claude can sort of see everything about the business. It's querying all those different sources?
If it needs to, yeah. It's literally like, if I say like, go and look up, you know, like the keywords, like if I wanted to optimize this landing page around like certain keywords, it would plug into data for SEO. If I said this page needs to be really fast, it would use like Lighthouse.
But yeah, every single one of these is plugged in and Claude can use it without. asking me permission because i've like configured the permissions in a safe way okay all right so it's going through everything it's seeing what you're doing creating a real infographic based on your work i would frankly just like to see that about my work just so i see what we're up to and and just see a visual of the way that i operate now that you've got the the uh the image how do you create the landing page and have it be meaningful right um i actually i'm just going to show you the the animation very quickly because it will take like two minutes So everything in this website, so this is my example landing page, right?
Everything's plugged into a GitHub repo. This is a Next .js website. All I need to do is say to Cloud Code, create a new version of this page.
And that might even be faster than creating the infographic because it's got everything it needs to basically duplicate this. and create a new one. So let's come back to Claude and let me just copy this.
Okay. So let's now come back to our, this was our growth agent, right? So our growth analyst that we were asking about performance.
So I'm going to say, create a new lead magnet page for this LinkedIn funnel infographic on the website. And when it's ready, show me a preview running on localhost.
So that is going to study the infographic, build out the page, run it on my computer first of all to kind of check that it works and then when I'm ready I'll say okay let's deploy this and then it will go up through Vercel so I can maybe just show you quickly my sort of tech stack. So Vercel is an incredible platform for just deploying code to the internet and I won't go into detail now but like it makes it so easy to like have an agent update your website everything's on github so that I can like jump back if I make a mistake everything's in something called sentry which is like error tracking so if something breaks I get an alert straight away I also have cubic which is like an AI code review tool so Claude Fable is pretty good, but we'll still make some mistakes.
And so Cubic will check everything before it goes on the live site and will often find like one or two issues. And then, yeah, I showed you Playwright already. This allows Claude to kind of inspect its own work and check that it looks good.
Okay. All right. Now, the landing page you are using is because PostHog told Claude that's the one that worked best.
Are there any tweaks based on how it... how it performed last time or based on what it's learned since last time?
Let's check. Let's see what it said. It's converting very well.
It's far ahead of some of your older pages.
It's saying it's a small sample.
It didn't actually pick anything up. I could ask it, you know, like, what can we do to further improve? like give me like three sort of like conversion rate optimization tweaks that we could make to this page um but already i've done quite a lot of optimization on the template so like there's a call to action here there's a call to action here there's another call to action here there's another form here like You just need to give your users as many opportunities as possible to click the button.
And the way that you know what to improve is Claude has access to data from PostHog. It sees where people are scrolling, where they stop scrolling, things like that, and it's using it. Got it.
Okay. All right. So I get the feedback loop.
What's next? Well, next is creating a new page for our LinkedIn lead magnet. And so Claude is, there's a document checklist.
So it's basically found something that the last agent sort of left behind, which is a checklist for creating new magnets. So this is just like a templated piece of work now. I don't need to reinvent the wheel.
Claude doesn't need to go and figure out how to do it. So it's just following instructions basically on building a lead magnet, which will probably take another... couple of minutes and then we'll have a web page we can preview.
In the meantime, I would be going into kit .com, which is where I run my email list. I absolutely love kit .com. It makes it, again, so easy for agents to analyze what's working, coding agents to build magnets into your site.
Obviously, it makes it very easy for humans to send broadcasts as well, and I can see nice stats on all my broadcasts. But what I would do is I would create a sequence. So here's an example sequence from the page I just showed you.
When someone fills in the form, they get this automatic email. Hi, Andrew. Here's the skill that you asked for.
Click here to download it. To set it up, just install it as a skill. And by the way, very subtle plug.
By the way, if you want to go further with agents, I run a four -week program. You can find out more here. So I'm not giving someone the hard sell.
I'm giving them what they came for, which is just like free value. And then I'm saying, by the way, if you want to know more, it's available here. And now that they're in my mailing list, I will send them an email when I release like a batch of new YouTube videos.
I will send them an email, for example, when I'm running a workshop. So tomorrow I'm doing a workshop on connecting email to Claude, which I think is risky. And I suggest people don't connect Gmail.
This went out to like a thousand people and a few hundred of them came from these lead magnets. So now they're like probably going to join a workshop. They've seen some good skills.
And then maybe if they're hungry to learn, they will join my sort of paid program. But I'm not like I'm not hitting them with the hard sell. Yeah, I get it.
All right. So this page is being built. It looks like it's nearly ready.
The other agent is just finishing off the animation. Here we go. Here's our nice animated infographic.
Now it's forced in this previous image, so we need to cut that. And I think this could be a little bit more eye -catching. If you imagine someone's looking at this on their mobile, I think it's a bit too text -heavy.
But you saw all I did was I gave Claude two prompts with Whisper. I answered a few questions, and now I have this infographic. and we can export it to a gif share it to linkedin and we would expect i would expect people to comment you know lead magnet to be given the skill yeah i totally see it and i also see that the the text on each one of those blocks should be a lot bigger at least the header text way bigger and i'm looking at a computer no i don't think it needs to get smaller each time i think each one should be bigger but you know this is like i said this is a sort of quick and dirty first draft But it's not bad and it's like a good starting point for then refining.
So, okay, cool. This is great. All checks pass.
So it's built the page. It's now going to start a server on my computer and it's going to check that this is working. So actually, if I just copy this, we should be able to test this ourselves.
Boom. Okay. It's taking a slightly different template because it doesn't have a big image right now.
But... Here's the form. Here are the things you get.
And then it's not quite finished adding this form. But if I needed a lead magnet very quickly, if I just needed somewhere to send people, I could deploy this. I could post the link to an infographic and people would start coming into the mailing list.
What's the connection into Kit? How are you doing it? Is it through API?
I missed that step. yeah so claude has a kit api key if i and it can actually and it can actually connect in and create its forms and then plug the forms into kit huh i had no idea yeah it can do everything so it can create tags so it would create a tag called like linkedin lead magnet it would create the form it would plug the form into this part of my site and then um it would basically yeah when someone joins it would drop them in all i have to do is create the sequence so the only manual part is like creating the email that like goes out people get after they sign up the nurture email yeah and again i want to i want to write that myself like i don't want to kind of send them ai slot but that's the only manual part that i need to do otherwise kit can sort of see and do everything yeah what do you use resend for that's another email service The resend is basically like transactional emails to sort of like when someone hits a checkpoint, they get an email from resend.
And that is just to that. The main reason actually is to protect the deliverability of kit. Okay.
That makes sense.
All right. So that's the process. This is how you think about it all.
And then writing is still you personally sitting down and writing. Anything else in here that we should see? Yeah, let me have a think about this.
I guess one thing I think is great is I use Atio as my CRM. And I've used this from day one. Really nice MCP, even better API.
And basically, there's a whole load of automatic CRM stuff happening in the background that Claude has created for me. If someone joins that... So if they join Kit, I could also create like a person for them in Atio.
I could then use Apollo .io to basically enrich them. So I actually do, when someone signs up, I get an email that says, someone just downloaded this skill. They're from a company that has 400 people.
That company does an estimated, you know, like X million dollars in annual revenue. And their title is, for example, like head of people. Now, when I see that email, I think, okay, this is my ICP.
They have the money to spend on AI training. The head of people is the person that's responsible for AI training. And so if I want to, I could sort of send them a manual reach out and say, hey, I saw you downloaded this skill.
What are you guys thinking about in terms of AI training for the second half of the year? That's all wired up with Claude automatically. And I also have an agent.
We haven't really talked about scheduled agents yet, but I've got an agent called Jamie. which is like James with AI. And basically, whilst I'm running like sales for my next cohort, I get a message every day at 8 a .m.
on Telegram. And Jamie basically tells me, here are the calls you have coming up. Here is the background on each of those people.
Here is the state of your pipeline. And the next cohort starts in, you know, like 55 days. here is like estimated revenue based on all the pipeline, like stages and probabilities.
So every morning I'm just getting like a little pulse check from Jamie that says you're roughly on target or like you might need to push a bit harder if you want to hit the target. And that is possible because it can see Atio, it can see Granola, it can see my to -do list, et cetera, et cetera. And that is Claude Code doing it in the cloud for those scheduled tasks, right?
This was, this was. Cloud Code running on a computer in my bedroom. So I set this up months ago.
I used something called the Cloud Code Telegram. This guy posted a great GitHub repo and he basically said like, this is a safe way to connect Cloud via Telegram. But what I use now, which I think is even better, is the Vercel Eve framework.
So they launched this like a month ago. basically allows you to run a single command in any folder and then it turns that folder into a cloud agent so you've got you know your context your code .md like all the files the agent needs you just say like start an agent now it's running in the cloud and you can run it across like any so you can use any model you can use like code models you can use like glm k3 whatever and then you can also connect any channel so like Slack, Telegram, WhatsApp for Business, email.
They've made all of those super simple to add. So now I use Eve by Vercel and I don't need to leave my computer running. I don't need to wire up cloud code to Telegram.
It just works in the cloud. When you say any folder, what do you mean by a folder? So when we're working with an agent, we often have a context folder, things it needs to know.
And this skill that I was showing was actually tidying up your folder, and making it really easy for agents to kind of find what they need. So in this example, I've got an agent called Advisor.
Cloud .md is just like defining the behavior of the Advisor. The more organized this folder is, the easier it is for an agent to find context. And the more organized this folder is, the less tokens you will spend trying to find the right information.
But still, tell me now about how to connect that and turn it into an agent. So all I would do is I would go into that folder in Cloud Code or in my terminal, sorry, and I would say like, you know, npx, whatever that, what that command was that they showed at the top, you know, show me my latest agent. But wait, show me that folder again, because what I want to understand now is the folder that you're showing the layout of that folder with cloud .md and so on.
Okay, let me just show you this. So I created a health agent that basically looks at all of my metrics, like Garmin, training peaks, and basically helps me with my triathlon coaching. And I used to just have this folder here.
This was all the data. But what I did was I said to Vercel, okay, let's create an agent. And so Vercel Eve has created a cloud .md, an agent folder.
you know, here are the channels I talked about, like Telegram. Here are some like scheduled tasks that it's running. So it's like telling me my metrics every day.
It's doing like an evening nudge. And so this is like, this is everything that like Claude needs to basically kind of like, oh, sorry, Vercel needs to run this agent. And then if I preview this, this is just like, it's very simple instructions saying this is a project using the E -framework, you know, go and look at these.
pages on the internet if you need help setting it up. And so this has all been packaged up. This is running in the cloud.
If I go to Telegram and then let's see if I've got web .telegram open. Again, editor might need to blur out. There shouldn't be any conversations here, but editor might need to blur something out.
Let me know if you want it. Yeah. Okay.
Where is my... Here. All right.
So it's been sending me an update every day on like my sleep, my recovery. So I can say like, how was my run last night? And I think it's actually, no, let's do first of all sync metrics.
So it's basically going to run a command in the cloud that like pulls all my sort of like data from my fitness wearables. So I can show you this here. Training peaks is what I use to like manage my workouts.
Heavy is when I go to the gym, like I show what I've lifted. Garmin. I let Claude see my sleep, my heart rate.
And I use this for health, but I also use this as part of my like weekly retro. So every Friday, Claude actually looks at my sleep as well and says like, you're telling me you feel stressed, but you also averaged like six hours sleep a night. No wonder you're stressed.
So it's like having a coach that kind of understands like your physiology as well as your psychology, which sounds like a small thing. But when Claude knows that you've had like not enough sleep, My Claude sometimes says, okay, I'm going to take what you said with a pinch of salt here because you are clearly sleep deprived.
So yeah, that's all plugged in. And there we go. Claude has just synced that.
So like, how was my run last night? So I went on a trail run yesterday up a mountain in Utah, which was quite fun. But we were at elevation.
We're like 1600 meters above sea level. And I noticed like it felt harder. So I think Claude might notice here.
that my heart rate was maybe higher, that my pace was slower. But we can see it's probably going to come back in a second. I found that with health data, Claude has a problem because there's so much data for it to go through that it has a hard time processing it.
I've structured mine in a certain way. I've like, I'll show you this. I've created a dashboard for Claude.
and i say i but i mean like claude has created this right so like yeah so this like organizes all my data into a single csv it's got like my weight my vo2 max my training readiness so i've basically structured the data so that it's easier for claude to find what it needs okay so here we go last night's session was your easy acclimation run uh you did eight and a half k in about 90 minutes average hr was 112 uh your tss was you know 36 versus the 31 that my coach had planned.
So I can now just, yeah, talk to this agent in the cloud. It will also nudge me. So I think like it was sending some nudges yesterday that I didn't read, but yeah, this is, this took, you know, one command that one sort of like command here in my folder.
And now there's like an agent running in the cloud doing stuff several times a day for me. All right. I do like this.
I hadn't tried Eve. I had no idea what it even did. I get it.
It basically, it just creates agents. It turns data sources into agents that I could then chat with. A hundred percent.
I think this could be like, I think this could be like the kind of open the claw killer or the Hermes killer. Too difficult for people to even understand. None of us even knew that it was, that it was there, let alone how it worked and what it did.
I think their main challenge is at the moment. Yeah. At the moment, this is kind of aimed at developers, even though it's like a, you know, it's a one command that you paste into a folder.
I still think they need to try and make this easier for non -developers. This stuff is super exciting. I love seeing what you're doing with this.
I've got one other founder that has done something similar using Tasklet. And I've got a link here for people who want to follow and see what that looks like. See you in the next video.
The Hook
The bait, then the rug-pull.
James McAulay runs a $250,000-a-month business with no employees, just Claude Code wired into every tool he owns. Andrew Warner has him rebuild one funnel live, from LinkedIn infographic to email sequence to CRM, to see exactly what 'connected to Claude' means in practice.
Frameworks
Named ideas worth stealing.
03:14list
The 3-step LinkedIn lead magnet funnel
Post a graphic that teaches
Gate the deep dive
Tag every signup at the door
A three-step content-to-email funnel: a LinkedIn post teaches something for free, a gated 'deep dive' collects the email, and every signup gets tagged so future posts target what they already showed interest in.
Steal forany single-post-to-email funnel
06:40concept
The LLM wiki (second brain)
Two to three thousand markdown files, organized by topic, holding eight months of business history so Claude can search it as plain text instead of the founder re-explaining context every session.
Steal forany solo operator using an AI agent as a long-term collaborator
21:23concept
Before/after agent context folder
A messy folder of duplicate, undated files reorganized into a fixed structure (CLAUDE.md, context/, briefings/, notes/, assets/, archive/) so an agent finds what it needs without burning tokens searching.
Steal forany agent context folder that's grown organically over months
25:37concept
Weekly Retro skill
A Friday-run skill that auto-pulls calendar, revenue, pipeline, and sleep/training data into one retro, cross-referencing stated stress levels against actual sleep hours.
Steal fora founder who wants one weekly check-in instead of five separate dashboards
Add Modern Creator as a preferred source and Google shows you more of our breakdowns in Search, Top Stories, and AI Overviews. It only changes what you see, and you can undo it in your Google settings anytime.
Add to Preferred SourcesOpens your Google source preferences with us pre-loaded. Tick the box and you're done.
Ben Senescu built OpenSEO to kill his own Semrush bill, wired it straight into Claude, and now watches open-source contributors fork it into CRMs and content tools he never planned to build.
Corey Ganim and Andrew Warner speed-run a week of AI launches, an AI QVC host, an endless AI livestream, and a phone number for your agent, and call out OpenClaw 2.0's rough demo along the way.
Andrew Warner sits down with Matt Pocock to unpack the Claude Code skill library that passed 200,000 GitHub stars, and the small habits of alignment, pruning, and teaching baked into it.
Andrew Warner spends a day cramming on "agent loops," then brings Matthew Berman on screen-share to react to five other creators' examples before showing off — and fixing — his own.
00:35
01:28
01:59
02:17
02:23
02:29
02:53
03:07
03:17
03:25
03:33
03:48
04:23
04:37
05:00
05:22
05:45
06:17
06:52
07:16
07:30
07:53
08:25
08:35
08:45
09:49
10:13
10:21
10:28
10:35
10:42
10:49
10:55
11:05
11:09
12:13
12:21
12:27
12:33
13:32
14:26
14:35
15:31
15:46
16:04
16:18
16:51
17:09
17:22
17:33
18:02
18:26
18:50
19:13
19:33
20:09
20:22
20:34
20:48
21:02
21:29
21:42
21:54
22:18
23:39
24:02
25:18
26:00
26:16
26:27
26:34
26:40
27:08
27:19
19:02
16:34
Video of the Day29:45
35:11
27:14
21:02