Modern Creator
Alex Cooper · YouTube

I Turned Claude Code into the Ultimate Creative Strategist

A creative strategist wires Claude Code to Slack, Notion, and an ad-research tool to build a self-updating brand brain that mines real reviews for ad copy and files its own reports.

Posted
2 days ago
Duration
Format
Tutorial
educational
Views
1.6K
73 likes
Big Idea

The argument in one line.

Connecting Claude Code to every tool in an ad account's stack turns it from a chatbot into a self-updating brand brain that mines real customer language for ad angles and delivers finished creative and reports on a schedule.

Who This Is For

Read if. Skip if.

READ IF YOU ARE…
  • You run or work inside a paid ad account and want AI help mining customer reviews and competitor ads for new creative angles.
  • You're a creative strategist or media buyer managing multiple client or brand accounts and want a persistent, self-updating context layer instead of re-briefing a chatbot every session.
  • You already use Claude Code and connect Slack, Notion, or Gmail daily as your team's system of record.
SKIP IF…
  • You don't run paid ads or track structured ad-account data — this workflow is built around an ad-account research tool, not general content ideation.
  • You're looking for a free or low-cost setup — the initial brand-brain build is described as costing a few hundred dollars in token usage on a premium plan.
TL;DR

The full version, fast.

A creative strategist clones a "brand brain" template into Claude Code and connects it to Parker (ad account and competitor data), Slack, Notion, Gmail, and an ad-generation MCP. The brain ingests years of Slack history, customer reviews, and ad performance data into a persistent local folder, then uses that context to generate static ad creative, competitor gap reports, and monthly performance decks from single prompts, all deliverable to Slack on a recurring schedule. The initial build is the expensive, slow step (hours, and reportedly several hundred dollars on a top-tier plan); ongoing use is cheap. The core habit: connect every tool as an MCP connector, ground generated ad copy in real customer language instead of letting the model invent it, and correct the system's reporting biases once so it remembers the fix.

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 →
Chapters

Where the time goes.

00:0000:41

01 · Cold open

Creator claims to have turned Claude Code into a 24/7 creative strategist that analyzes the ad account, finds ad inspiration, and creates ads without manual prompting.

00:4102:09

02 · What Claude Code Actually Is

Explains that Claude Code keeps context in a local folder instead of a chat log, so it compounds knowledge across sessions in a way regular Claude or ChatGPT chat doesn't.

02:0904:05

03 · The Tools You Need To Start

Walks through Settings > Connectors, adding Calendar, Gmail, Higgsfield, Notion, Slack, Gamma, and a custom Parker MCP connector for ad-account data.

04:0507:08

04 · Setting up the Parker Brain MCP

Clones the parker-brain GitHub template into Claude Code; explains the three-phase build (learn brand DNA, decide a point of view, make the work) and that the initial build costs a few hours and, on a top-tier plan, a few hundred dollars in tokens.

07:0811:11

05 · The Parker Brain in Action

Loads the generated brand folder, runs onboarding, then asks Claude to read the entire client Slack channel history and fold every piece of client feedback into its context.

11:1113:23

06 · Connecting to Slack

Claude finishes reading three years of Slack history into a consolidated feedback document, then the video shows the resulting Routines list and a Notion ideas library populated automatically from that context.

13:2316:56

07 · Building Creative Reports

A single prompt generates a full Gamma deck covering competitor moves and gaps, customer voice in their exact words, and untested audience openings, plus a Google Slides breakdown of a competitor's concentrated-offer strategy.

16:5618:21

08 · Static Ads Workflow

Prompts Claude to mine customer reviews and ad comments for emotionally loaded phrases, cluster them into angles, and generate finished static ads via the Higgsfield MCP.

18:2118:53

09 · Setting up Routines and Schedules

Shows converting the static-ad workflow into a recurring routine that runs weekly and delivers new ads straight into Slack.

18:5319:56

10 · Setting up Parker Correctly

Stresses that output quality depends on correctly tagging competitor, affinity, and inspiration brands and enabling AI tagging on every ad inside Parker.

19:5620:45

11 · Summary

Closes with a link to a free month of Parker and a reminder that the workflow requires generating the brand brain first.

Atomic Insights

Lines worth screenshotting.

  • Claude Code retains context in a local folder instead of a chat log, so knowledge compounds across sessions instead of resetting each time.
  • Connecting every tool you use as an MCP connector is what turns a single AI chat into a system that can act across your whole stack.
  • A brand brain built from a GitHub template can ingest years of Slack history, ad account data, and customer reviews into one versioned folder that gets smarter every time it's queried.
  • Generating an initial brand brain is the most token-intensive task in the workflow: expect a multi-hour run and, on a premium plan, several hundred dollars in usage.
  • Once the brain is built, day-to-day usage no longer needs the top-tier plan; only the initial ingestion does.
  • Scheduled routines let the brain refresh its own context and generate new ad ideas automatically, without a person re-prompting it.
  • Feeding real customer reviews and ad comments into the ad-generation prompt, rather than letting the model invent copy, is what keeps output from reading as fabricated.
  • A single prompt asking for a monthly creative performance report can produce a finished slide deck covering competitor gaps, voice-of-customer findings, and untested audience segments.
  • Correcting a model's reporting bias is a one-time plain-language instruction, and the brain retains that preference for every future report.
  • Static ad generation from real reviews needs almost no human editing pass; video ad generation still needs a human in the loop.
  • The value of a system like this scales with how well the underlying data (tagging, categorization, historical ingestion) is maintained, not with the AI model alone.
  • Turning any one-off Claude Code output into a recurring Slack delivery is just asking the model to "turn this into a routine" — no separate automation tool required.
Takeaway

Feed the AI real customer words, not invented ones.

WHAT TO LEARN

The most durable lesson isn't the tool stack, it's that ad copy grounded in verbatim customer reviews and comments beats AI-invented copy, and a persistent context folder makes that grounding compound over time instead of resetting every session.

02What Claude Code Actually Is
  • Persistent local context beats a chat history: information saved to a folder compounds across sessions instead of being re-explained every time.
03The Tools You Need To Start
  • An AI assistant's usefulness scales with how many of your actual tools it can read and act inside, not with the model alone.
04Setting up the Parker Brain MCP
  • Building a comprehensive knowledge base from scratch is the slowest, most expensive step in any AI workflow — budget real hours and real token cost for it once, then work cheaply afterward.
  • A higher-cost model produces a deeper first pass; a cheaper model is fine once the base context already exists.
05The Parker Brain in Action
  • A short onboarding conversation about objectives and key metrics at setup time is what lets later, shorter prompts stay accurate.
  • Ingesting a full multi-year history in one pass turns scattered feedback into a single reference document every future request can draw on.
06Connecting to Slack
  • Letting the system write its own summary documents, rather than an analyst compiling them by hand, keeps them current automatically.
07Building Creative Reports
  • A single well-scoped prompt covering what's working, what's not, grounded in customer evidence, can produce a full report deck that would otherwise take days to assemble by hand.
  • Comparing your own account against tagged competitor and affinity-brand data is what surfaces untested angles, not brainstorming in a vacuum.
08Static Ads Workflow
  • Grounding generated ad copy in verbatim reviews and comments, rather than letting the model invent language, is what makes the output usable with minimal editing.
  • Static creative comes back close to ready-to-run; video creative still needs a human pass.
09Setting up Routines and Schedules
  • Any one-off output can become a recurring, scheduled delivery just by asking for it, no separate automation tool needed.
10Setting up Parker Correctly
  • The output quality of a system like this is capped by how well the underlying data (tagging, categorization, historical ingestion) is maintained, not by the AI itself.
  • Correcting a system's bias, like over-indexing one metric, is a one-time instruction whose effect persists in every future output.
Glossary

Terms worth knowing.

MCP
A connector standard that lets Claude Code read and act inside other apps and data sources, like Slack, Notion, or an ad platform, instead of only chatting in a text window.
Brand brain
A versioned folder of context Claude Code builds and updates about one brand or client, including audits, personas, competitor notes, and running feedback, that every future session reads from.
Routine
A workflow inside Claude Code set to run on a schedule, such as every Monday, instead of being manually triggered each time.
Static ad
A single-image ad creative, as opposed to a video ad, which is faster and cheaper to generate and test.
Voice of customer
Research drawn from actual customer reviews, comments, and support messages, used to source language and angles for ad copy instead of inventing it.
Whitespace
An audience segment, angle, or format competitors haven't tested yet, surfaced by comparing your ad account to competitors'.
Resources

Things they pointed at.

02:58toolHiggsfield MCP
03:07toolGamma
03:14toolNotion
03:05toolSlack
Quotables

Lines you could clip.

00:00
I turned Claude Code into a superhuman creative strategist, and now it works for me twenty four seven analyzing my ad account, finding me new ad inspiration, and even creating me ads all without me lifting a finger.
cold-open thesis statement, no setup neededTikTok hook↗ Tweet quote
02:05
It's like it really is like co work on steroids is is Claude code.
punchy comparison to a competitor AI toolIG reel cold open↗ Tweet quote
14:32
The most liked comment in the last sixty days is he's like my 16 year old newborn. He's talking about dog sex.
funny, specific real customer-comment example that breaks the tutorial toneTikTok hook↗ Tweet quote
14:54
The world really is your oyster when you get into cold code and you get into producing things like this.
aspirational closing line for the workflow sectionIG reel cold open↗ Tweet quote
16:50
This folder gets smarter, and therefore, your Claude gets smarter the more that you use it.
tight thesis line summarizing the whole systemnewsletter pull-quote↗ 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.

metaphoranalogy
00:00I turned Claude Code into a superhuman creative strategist, and now it works for me twenty four seven analyzing my ad account, finding me new ad inspiration, and even creating me ads all without me lifting a finger. On top of that, I built a living self improving brain that gets smarter every time me or my team use it. And today, I'm gonna teach you exactly how to do that even if you've never used ClaudeCode before.
00:20I truly do believe that the genetic coding tools like ClaudeCode and Codex are the biggest opportunity for Creative Strashes today to multiply themselves and make themselves incredibly more productive with AI. So today, I'm going to be walking through what they actually are, how to install the right tools to make sure everything's set up the correct way, and then some use cases of exactly how I use Code to make myself a 10x marketer and make more and better ads.
00:41So let's start by talking about what ClaudeCode actually is. See, when you use the regular chat, you are just beholden to the memory of ClaudeCode or ChatQPT chat remembering everything that you say.
00:54The cool thing about using ClaudeCode or Codex is that the memory or context is actually baked into your local machine like this, which means it's much better at retaining information, and then you can compound this knowledge a lot easier than you can in the chat, which means it does, by nature, get smarter every time that you use it.
01:13Instead of having a chat in Claude AI, you're actually chatting inside of a folder that has all of the context about that specific project. And you can do all the same things in here that you can do in the normal Claude chat. You can have MCPs in here that generate visuals or do reports or make artifacts.
01:28All things you can do inside of Claude chat, you can do inside of Claude code. And because of the fact that it compounds over time is the more you use it, once you start using this stuff, there actually is no reason to ever go back to normal Claude chat because Claude code beats it in in every way possible. Sometimes I use co work, but like the vast majority of my chats now are inside of Claude code.
01:47It can be a bit intimidating when you first get started, but honestly, once you understand the file structure and you start having conversation here, you will pick it up and become second nature.
01:56And I assume that you'll have the same kind of epiphany and and not ever go back to the the regular Claude chat.
02:03It's like it really is like co work on steroids is is Claude code. Anyway, let's talk about the tools that you need to make this all possible.
02:13So what we're gonna do is gonna go to customize and then connectors. And what I want you to do is spend some time thinking about all of the tools that you use in your ad creative workflow currently or just in your in your, you know, whole tech stack and add as many of them to Claude as connectors as possible.
02:31The reason for this is because Claude code is incredibly powerful on its own. When you pull in all of these other tools and have it work inside of Slack, inside of Notion, inside of, you know, your email, that's when it gets really, really powerful.
02:46So you wanna go and connect all of these as soon as possible. I have a Google Calendar so I can see all my meeting notes, Gmail, obviously.
02:54The Higgs field MCP is very useful because that means we can make statics and videos with it. Notion because we do all of our briefings out of Notion.
03:03Slack because I wanted to ingest all of the context from Slack. And that is most of them that we'll want to have Gamma because we do a lot of our reporting through this. So that's also useful as well.
03:16And what you will need for today's video is the Parker MCP. That's how we're gonna pull in all of the information from the ad account and other ad libraries, AI tags, etcetera. This is gonna make this a lot lot easier to do.
03:28You will need a parker account for that if you want to give us a try for free. I'm gonna leave a a link in the description below to get a free month of parker and, uh, try this. All you're gonna wanna do is grab this link here that I'll put in the description.
03:39Again, go to customize and connectors. Go to add a custom connector, and then just call it Parker.
03:51And then you're gonna put in the URL and click add, and you'll have the Parker MCP inside of here. That's gonna pull all of your ad account data into here.
03:58It's gonna pull data about professors, pulling all your tags, etcetera. There's a lot that comes with that that's gonna make this process a lot more powerful. The next thing we're gonna wanna do is pull in the Parker brain.
04:08Now, this is the part that makes it self improving, pulls in all the information from your Slack, from your Notion, or wherever you do work, and actually updates the context in your sleep. What we're gonna want to do is go to this link, which I will link in the description, copy the link, bring it back inside of our Claude code here, and we're just gonna say in a folder that you can create, you know, I've just got a general Claude code folder created.
04:30You can just say, you know, I want you to clone this brain for my brand x.
04:41Follow the instructions exactly as is, and you're gonna paste in the link. And then you're just gonna hit enter. You can do this on Fable five.
04:48It is going to cost you a few $100 if you do that because what's basically gonna happen now is it is going to run a ton of reports and audits and generate a lot of context on your brand that saves to your folder that you wanna work inside of. And the initial creation of this is a lot of work, and it'll take a few hours to go and do this.
05:08So I would recommend putting on auto mode so you don't have to keep accepting, you know, permissions and edits. But, yeah, just know it will take a bit of time whether you use Fable or Opus.
05:21If you wanna save money, then I would do it on Opus. If you want to, you just make the absolute best brain possible, then I would do it on Fable.
05:28But just know it will take a few hours to generate, and the Fable one will cost you a few $100. I would also strongly recommend having a max plan just to generate this. If you have a max five x plan on Claude, you will likely run into five hour token limits once or twice.
05:40If you have the 20 x plan, you're good to go, and you should be able to generate these with no issues. But that is gonna be by far the most token heavy thing that we do here.
05:50I mean, the reason that it it takes so many tokens is because it's learning and going so deep on your brand to generate context that is going to make the outputs way smarter, which is why this stuff is important. But as soon as you generate the initial, you know, batch of context, then, you know, you can downgrade because you won't you need to be on a 20 x plan.
06:07So what's gonna happen after you've done that? It's just gonna ask you a few simple questions in a pop up here. If I go to an actual brand setup that I did yesterday, you know, it's just gonna ask, you know, what's your main objective, North Star metrics, just just to learn about your, you know, your setup, and then it's gonna go and generate this.
06:25And like I said, it'll take a few hours. But once it's done, you are ready to get started and making some of these really incredible workflows. So the only thing you're wanna do is just open up a new chat and then go to load up the brain.
06:39So the one that I just generated there was my overpop client brain. Just gonna open that, trust the workspace, and now we're working inside of this overpop brain.
06:50I'm just gonna do the command get started, and that's gonna walk me through, again, a little onboarding here that's gonna make sure that all of my routines are set up.
06:59We'll get on to what they are in a moment, and everything is in the right place. And you train the brain on exactly what it needs to know, and then we can start getting into the fun stuff. So let's actually see in action.
07:10What could we do here? So I could say, for example, I want you to look through my client channel, add a create Opapop, and ingest all of the feedback the client has ever given and include this in your context going forward.
07:37Now because we have set up Slack as a connector, it's going to look through the entire channel's history, and any feedback the client has ever given is going to bake it into the context here. And going forward, anytime that the client gives feedback inside of Slack, we know that that is gonna flow through to our client brain.
07:54And every time I query inside of Claude code, it will already have that context. Just gonna click always allow and allow it to run this again. I would recommend putting it on auto mode so you don't have to keep accepting permissions.
08:06And this is why it's important to connect as many tools as you use with the MCPs into Claude because then it can ingest all of this context, and it can start outputting. You can get to start sending messages inside of Slack. You can get it to start adding things to Notion.
08:19For example, I was in a Claude chat the other day where I was asking it to come up with ad ideas every week, and then input them into my idea library inside of Notion. This is literally the exact idea library that we use. And Claude, via the Parker MCP, was coming up with ad ideas and adding them into here completely automatically, and then adding links to, you know, TikToks it was finding or comments that it was referencing from my ads.
08:44And this all comes because I've got my Parker set up in the back end to know all of these things. So I have the customer reviews uploaded here.
08:53I have post purchase surveys. My ad comments flow in automatically inside of here. It scrolls TikTok for me and finds TikTok inspiration.
08:59Ad library, I've got competitors that I follow, so it knows what my competitors are. So all of the context that you wanted to know, I just basically look at the the web app here as a as a control room. And, if you wanna try this, there's a there's a link to a free month in the in the description down below.
09:14And then all of this information flows through the MCP into Claude so I can get it to do these workflows where it says, you know, look at my competitors and everything they've launched in the last thirty days and, you know, look at what ad formats they've run, find their top ads by impressions, and then go and remake them for us using the Higgs field MCP.
09:37Like, those are the flows that become possible when you have this Parker MCP that's pulling in all the data about your ad account, your your competitors, and then connecting that to the Parker brain that has all this context that self improves over time.
09:53And then actually connects to the tools that you use on a day to day basis so it comes to you rather than just, you know, you chatting inside of cloud all the time. Now going back to this chat where it's ingesting the Slack context, you can see it says I've read the full ad create OpPop history from June 2023 to today, and it is now writing into its context.
10:13So it's gone into Slack, ingested all of the messages from the last three years, and now it is writing that into the folder so that all future queries that you do already has context of not only the brain, the ad account, everything inside of Parker, but also every message ever in Slack. You can do this for Notion as well, every concept we ever produce.
10:32You can do this for, you know, Google Calendar for every meeting that you've had with the client, for every internal brainstorm you've ever done. Like, all of this context should be going into your client brain so they get smarter every day. And this is where it becomes really interesting because, you know, this is already accounted for if you have the Parker MCP.
10:48We do this automatically for you. We set up these routines for you, but you can start setting up routines that do this on a schedule. So I go into my routines here.
10:58I can see, you know, every day, we've got overpop, you know, update in the brain or, like, refreshing the context or generating new ideas for me. So all of this stuff is automatically being updated, so it's getting smarter and smarter over time.
11:11Then when you create any workflows inside of here, for example, I was creating a workflow here where I said, look at my external brands top static ads by impressions. Again, that is from Parker.
11:24So look at my external brands. Look at their top ads by impressions top static ads by impressions, and find ones that it makes sense for us to recreate.
11:33Draft a copy to recreate them. We would then do through the Hicksville MCP. It went and generated these statics.
11:42And, yeah, I think the Hicksville MCP's playing up here. It sometimes seems to be a little bit inconsistent with, like, showing on on old chats. But just to give you an idea, I had it delivered them to me in Slack and I was pretty impressed by how well these came out.
12:00I mean, these a lot of these I would say I'd genuinely good to go inside the ad account. I mean, we could do a little bit of work on the copy because I literally didn't do any work on the copy of this.
12:10So just was just testing this the other day. But you can get ads delivered into your Slack or wherever you operate on a Monday morning just by setting up a workflow like this and then saying, okay.
12:24Now turn this into a routine that runs every morning and pings me inside of my internal Slack channel. And again, these because it's connected to the Parker MCP are real comments or reviews that our customers have have said.
12:41So it's not like fabricating this is not getting them for it's not just making the opposite itself. I ask go and find a real customer reviews or real ad comments or real post purchase surveys, and then infuse them into ads that have performed for external brands and and remake them for us, then deliver that to me every Monday morning.
13:01You could have it make 25 ads for you and send it straight into Slack every Monday morning using the Hix Steel MCP, which again is another one that you wanna set set up here inside of your Claude. And that's when this just becomes so incredibly powerful when you think about all the different tools you have and then just start connecting them together and having it proactively deliver them to you on a schedule inside of the place that you already work.
13:23You know, mean, spoke about Gamma earlier. Another report by another thing that I did was I said, build my monthly creative port report. Again, this is, like, a super simple problem.
13:33You can probably make something way, way better than this. I just want a performance story over time. Tell me what's working, what's not working, and then ground it in customer evidence.
13:43And I made it make that inside of Gamma, and here's what it did. So it's literally put together this whole report with everything that I need.
13:51You know, we add, create, do retros every month. We have to produce a report just like this and then send it to the team internally, and then send to the client. This has been done in one prompt with no optimization.
14:02I could say, you know, go and, you know, index more on competitors or go and do more organic research. Because you are tapped into everything through the Parker MCP.
14:11You can get to report on performance. You can get to report on competitors. You can get it to look at, you know, our customer voice and compare it against the ad account and see what opportunities we're missing.
14:21You can get to look at the organic feed. You can get to look at ad comments. For example, it says here, the most liked comment in the last sixty days is he's like my 16 year old newborn.
14:29He's talking about dog sex. This is a dog brand. That's a really, really nice piece of copy and an angle that we actually haven't tested yet, which flagged inside of the ad account.
14:39It got four to six likes as well as a common most like common in the last sixty days. So that is a a good signal that we might wanna start making ads around senior dogs. And all of this stuff is just put together in this one report that we've made through one prompt, which is pretty insane.
14:54Again, here, it's gone on the organic feed and found content that we might wanna recreate. It's looked at our other brands inside of Parker and and looked at what they're testing right now and and and given us suggestions. There are so many things that you could go and build off the back of this.
15:11The world really is your oyster when you get into cold code and you get into producing things like this. Here's another report that I made in Google Slides, another one that looked at a competitor of ours and compared us to the gap analysis and then used that to come up with ad ideas based on our voice of customer.
15:27Imagine having a v one of your monthly report that you send to, you know, your performance team, your strategy team, your client, whatever it is, delivered to you inside of Slack every single month saying, here's the b one of the report. Approve it, and I'll send it off to the relevant stakeholders or go and edit inside of Gamma or give me feedback, and I'll know for next time.
15:46And that's the cool thing here is you can literally if I didn't like this for whatever reason, I could go into Claude code, and I can say, when you're doing reporting, you seem to over index on, you know, ROAS or CPA. I care more about the ads to get the most spend inside the ad accounts. From now on, I want you to remember this and in all future reports that you do, always prioritize that metric.
16:07And then the brain will take that feedback on board, write it into its system, and then know for next time that that's how you do reporting. If we go back to the original query that we had about getting the Slack context updated, you will see that it's now completed that.
16:22It has wrote it into the folder. If I go into the folder, this is the open pop folder that we made. I can see that if I go into running notes that said here, we can see it's got a client feedback document.
16:34It has created this document. I had not created this document. It has taken in my query and created the client feedback document.
16:40It's going to update that as more context comes in from Slack, comes in from Google Calendar, comes in from email or wherever you wanna connect, and this system gets smarter. This folder gets smarter, and therefore, your Claude gets smarter the more that you use it.
16:55I do just want to highlight the effectiveness of the static ads workflow again. This has been one of my favorite use cases of the Parker MCP and the Parker Brain so far and Claude code in general. Uh, because here I said, you know, search my customer reviews and face value comments, find the most emotionally loaded phrases that are that our customers are telling us, uh, cluster them into distinct angles, look at the Facebook ads data, and then basically use these to create some static for us.
17:20And then using the Hickfield MCP, it generated these, which again, I have not, you know, done a pass on the copy here. But this is, you know, this is the phrase that we saw from the report, so it knows, you know, where our white space is, where we haven't tested inside the ad account yet, what personas we have not here.
17:34And I can tell you because we do the ads for this account, we have not really hit the senior dog angle at that high yet. Kinda that is a that is a ready to go static that I would launch inside the ad account.
17:45That's very good. And this is a real customer comment or review that they have set, so we can use that and put it into the account. So just just setting up this workflow and getting these delivered into your Slack every Monday, it's just such a lock, and you can put out so many more statics.
18:02You can do it for video as well, but you need a little bit more human loop for that, I found. But statics, it's just such like, statics and reporting, it's so so good for this and getting it to populate your ideas bank like we showed it in in Notion, whether you have yours somewhere else, you can do that too. The amount of the amount of workflows you can just set up off the back of this are, you know, absolutely insane.
18:21And again, anything you want to turn into a schedule, you can just say turn this into a routine, cut spell, and send this to me on Slack every Monday.
18:32Or you can just go and set up the routine in the routine section, and you can have it here. But, you know, if you use the parka brain, then we're gonna set up a a bunch of these routines for you.
18:42So it will self update. It will dream. It'll refresh the context.
18:44It'll come up with ideas for you every single week. Create research loops. Loops and judges is a whole another video I can get into as well.
18:51There is so, so much that you can get into here. And as a reminder, the outputs will be as good as you have your Parker set up on the back end. So make sure you're following brands inside of here.
18:58They're tagged as, you know, affinity brands, which are, you know, brands that are, you know, sent to the same audience but not the same product, competitors, inspo brands. Make sure you have those set up.
19:08You've got AI tagging on every single one of the ads, even your competitor's ads. So you can ask, you know, look at Caraway's ads and give me a breakdown by ad format. Every ad in our database is ranked by impressions, so you can kind of see what's performing for other brands.
19:22You can say, you know, look at the Caraway's top ads by impressions and try and find some common themes in, you know, the emotions or desires or ad formats or just in general. You got the same thing in the performance tab. You can say, look at my swipe file.
19:35Look at my ideas. Basically, anything that you can query inside of Parker, you can query inside of the Claude MCP. Now Reddit, do not have access to yet.
19:44We are working on that. It's close, so I'm very excited to to release that to you guys. But everything else inside of the app, um, you can you can query inside of the MCP, um, and get it to turn into ads for your brand.
19:56So if you want to give this whole workflow a try, I'm gonna give you a link to get a free month of Parker. To try completely free, all you would need to do is generate the brain inside of your Claude code. So click the link in the description below.
20:09You can get a free month of Parker. Try this completely risk free and really just step into this new world.
20:15I'm so excited for you guys because genuinely, once you start using Claude code or codex, there is no reason to go back to regular chat. There is no reason to ever have a chat in Claude AI again once you've got these brains set up for your brand.
20:29And I'm so excited about to see the use cases that you come out with because I'm sure there's a bunch of other things I've even, you know, thought of that you guys can do with this, with all the different tools you're gonna connect. Anyway, if you like this video, hit subscribe, comment what you wanna see more of, and I will see you in the next one.
The Hook

The bait, then the rug-pull.

Claude Code isn't just a coding tool: plugged into Slack, Notion, and an ad-research platform, it becomes a standing brand strategist that reads every customer review, every competitor ad, and every piece of client feedback ever left in a channel, then turns it into new ad creative and reports without a fresh prompt each time.

Frameworks

Named ideas worth stealing.

05:50list

The Three-Phase Brand Brain Build

  1. Parker learns your brand DNA — audits the ad account, reads reviews, builds personas, profiles competitors
  2. Parker decides a point of view — the strategy: who to target, what to say, what to make
  3. Parker makes the work — ideas and briefs built off that strategy

The three-step process the brand brain runs once cloned: audit and learn, decide a strategic point of view, then produce creative and briefs off that strategy.

Steal forstructuring any AI-assisted creative-strategy build-out
15:20model

One Machine, Pointed at One Nerve

A Google Slides framework built inside the report, arguing a competitor runs one emotional message at volume behind one hard offer, while a varied portfolio spreads the same emotion thin and under-funds the offer.

Steal forauditing whether an ad account's creative is spread across too many concepts instead of concentrated on one
CTA Breakdown

How they asked for the click.

VERBAL ASK
19:56link
I'm gonna give you a link to get a free month of Parker. To try completely free, all you would need to do is generate the brain inside of your Claude code.

Soft-pitched once early (around 3:07) then made the explicit ask only in the final 45 seconds, after the full workflow value has already been demonstrated.

MENTIONED ON CAMERA
Storyboard

Visual structure at a glance.

open
hookopen00:00
connectors
valueconnectors02:15
clone brain
valueclone brain06:37
slack ingestion
valueslack ingestion10:55
gamma report
valuegamma report13:25
CTA
ctaCTA19:06
Frame Gallery

Visual moments.

Watch next

More from this channel + related breakdowns.

Chat about this