12 Claude Plugins, Skills & MCPs I Can't Live Without
A rapid tour of twelve CLIs, plugins, and connectors that extend Claude Code and Claude Cowork past their default integrations, from a Gmail connector that can't send mail to a scraper that can.
Posted
4 months ago
Duration
Format
Listicle
educational
Views
24.4K
609 likes
57 · 43
Big Idea
The argument in one line.
Claude Code's default connectors and MCPs are often read-only or token-heavy, so a stack of purpose-built CLIs covering Google Workspace, design, scraping, and outreach does more for less and should replace most default integrations.
Who This Is For
Read if. Skip if.
READ IF YOU ARE…
You use Claude Code or Claude Cowork regularly and want to extend it with external tools instead of staying inside the default connectors.
You're already paying real money in tokens each month and want concrete ways to cut that bill.
You do outreach, lead gen, or web scraping and want AI-native tools instead of a legacy SaaS dashboard.
SKIP IF…
You've never used Claude Code or a CLI before; several of these plugins assume you already know how to install and run one.
You want one deep dive instead of a 12-item tour; each tool here gets only a few minutes.
TL;DR
The full version, fast.
This video runs through twelve tools that extend Claude Code and Claude Cowork beyond the default connectors: a Google Workspace CLI that can send mail and edit sheets where the stock Gmail connector can't, design tools (Higgsfield, Impeccable), instant HTML hosting on Vercel, a 'Printing Press' plugin that builds custom CLIs from any API, a token-compression tool called Caveman, scraping and browser automation via Firecrawl and Playwright, a video-watching plugin for visual analysis, and two outreach tools, Vibe Prospecting and Unipile. The throughline: a narrow CLI built for one job usually beats a general-purpose connector on both cost and capability.
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.
States the premise: Claude Code and Cowork have a huge ecosystem of skills, plugins, MCPs, and CLIs, and this video picks the top 12 worth using.
00:22 – 03:24
02 · Google Workspace CLI
A CLI that connects Drive, Gmail, Calendar, Sheets, Docs, and Chat with none of the read-only limits of the default Google MCP; demoed building a sponsorship-tracking sheet from inbox search.
03:24 – 06:22
03 · Higgsfield
An image/video generation tool (Nano Banana, GPT Image 2, video models) usable inside Claude via MCP or CLI, with reference-image consistency and the ability to wrap it in a reusable thumbnail-design skill.
06:22 – 08:15
04 · Printing Press
A plugin with 50+ pre-built CLIs for API-less tools (ElevenLabs, OpenRouter, Shopify, etc.), plus the ability to build a custom CLI for any MCP or even for software with no public API at all.
08:15 – 10:04
05 · Impeccable
A set of skills for click-to-select editing of generated HTML and websites: adjust layout, animate, polish, or check against a brand guideline file.
10:04 – 11:20
06 · Vercel
Deploys any generated HTML to a live URL in about 10 seconds, used for client onboarding pages, dashboards, and sales proposals that can be updated after they're sent.
11:20 – 13:35
07 · Caveman
Compresses frequently re-read text (CLAUDE.md, skills, MCP docs) to cut token spend; a real CLAUDE.md got 40% shorter and a skill lost 18% of its input tokens with no quality drop.
13:35 – 15:26
08 · Firecrawl
A scraping tool that succeeds where Claude's native web-fetch tool fails on JavaScript-heavy sites, returning clean Markdown instead of raw HTML; in a 5-site test it scraped 4 sites the native tool couldn't touch.
15:26 – 16:58
09 · Playwright CLI
A browser-automation library that takes actions through generated Python scripts instead of visual reasoning, making it faster and more reliable than Claude's native browser-use for repeatable site workflows like form-filling.
16:58 – 18:15
10 · Cloud video plugin
Watches a video by downloading it and taking screenshots, not just pulling a transcript, enabling visual breakdowns like a competitor-reel analysis sheet with hook screenshots.
18:15 – 19:35
11 · Vibe Prospecting
An AI-native lead database that filters on live intent signals like hiring activity or funding rounds, not just static firmographics, and shows query cost upfront.
19:35 – 21:41
12 · Unipile
Connects Claude to WhatsApp (via an unofficial API and QR login), Instagram, and LinkedIn for reading and sending messages; closes with a subscribe ask and an AI Accelerator pitch.
Atomic Insights
Lines worth screenshotting.
MCP connectors load their entire API schema into every new chat, while CLIs only load on demand, so swapping an MCP for a CLI can cut token spend noticeably.
The default Gmail connector is read-only: it can't send email, create a calendar event, or edit a Google Doc, but a Google Workspace CLI can do all three.
Compressing a CLAUDE.md file to a 'caveman' style cut one real file by 40% with no loss of function, which matters because that file gets re-read every single session.
A business running 30 Claude Code sessions a day burns through roughly 900 sessions a month, so trimming a repeatedly-read instructions file compounds fast.
Running the same compression pass on a skill cut 18% of its input tokens with no measurable drop in output quality.
Claude's native web-fetch tool fails on JavaScript-rendered sites; in a five-site test, a dedicated scraper succeeded on four of the five sites the native tool couldn't touch.
Browser automation driven by generated Python scripts is faster and less error-prone than visually re-reasoning about a page on every step, because the script doesn't have to re-find the button each time.
A video-analysis plugin that downloads the file and takes screenshots gives visual context a transcript alone can't, like which exact frame made a thumbnail or hook work.
Lead databases built for AI agents can filter by live intent signals, like a company's current hiring activity or a recent funding round, not just static filters like industry and headcount.
Reusing a personal WhatsApp number via QR-code login skips the official WhatsApp Business API approval process entirely, but the account's own platform limits still apply in full.
Outreach automation platforms cap daily actions (the video cites roughly 20 a day on LinkedIn), a limit worth confirming directly with the platform rather than trusting a secondhand number.
Takeaway
Replace default connectors with purpose-built CLIs
TOOL STACK AUDIT
Across twelve tools the throughline is the same: a narrow, purpose-built CLI usually beats a general connector on cost, reliability, and what it's actually allowed to do.
02Google Workspace CLI
Default connectors are often read-only by design: a Gmail connector that can't send mail or a calendar tool that can't create events is a deliberate limitation, so check what a connector can actually DO before assuming it covers a workflow.
A CLI and an MCP both sit on top of the same underlying API, so the choice between them is a cost-and-capability tradeoff, not a feature tradeoff: pick whichever surfaces the specific actions the job needs.
03Higgsfield
Feeding a reference image and a brand guide into an image tool makes repeat outputs consistent, which matters more for recurring design work than any single generation's quality.
Turning a one-off generation into a reusable skill is what actually compounds: the second and hundredth use get faster because the setup questions are already answered.
04Printing Press
A library of pre-built integrations for tools with no public API closes a real gap: most 'no API' tools still have internal endpoints a determined wrapper can reach.
Building a one-off integration from scratch for a single frequently-used tool pays for itself the moment it replaces a slower, more error-prone manual process.
05Impeccable
Visual, click-to-select editing of a generated interface closes the loop between describing what's wrong and seeing it fixed faster than describing changes in prose.
Codifying brand guidelines into a reference document a tool can check designs against turns subjective design feedback into a repeatable, automatable check.
06Vercel
Static deliverables like PDFs and slide decks can't be updated after they're sent; a live, hosted page can be revised after the fact without resending a new file.
A document built in a format a model already works in natively is cheaper to iterate on than one that has to be translated into a separate export format first.
07Caveman
A large instructions file that gets re-read on every single session is one of the most compounding costs in a repeated workflow: a fixed per-session overhead multiplied by every session, every day, adds up fast at scale.
Compression of standing instructions is worth testing against the original for a quality regression before trusting it, not after: a shorter file that quietly breaks functionality is a worse trade than no compression at all.
The same compression principle applies to any large, frequently-reused reference document: the savings scale with how often the text gets re-read, not with how big it is.
08Firecrawl
A tool's 'it can read the web' claim and 'it can read THIS site' are different claims: JavaScript-rendered pages routinely defeat simple fetch-based tools even when they succeed on static pages.
Getting content back in a clean, pre-structured format instead of raw HTML cuts the processing cost of every downstream step, not just the fetch itself.
09Playwright CLI
A tool that takes direct, scripted actions on a page is more reliable than one that has to visually re-interpret the layout on every step, because the script doesn't have to re-decide where the button is each time.
A scripted workflow built for one specific site can be reused as-is next time that same site needs the same action, where a one-off visual pass starts from zero every time.
10Cloud video plugin
Analyzing a video from its transcript alone misses the part of the message that was never spoken: the framing, the on-screen text, the exact frame where attention was captured.
Combining a visual capture with a written summary turns a vague note like 'this one worked' into a specific, checkable claim about which frame or moment did the work.
11Vibe Prospecting
Static firmographic filters describe who a company IS; behavioral and intent filters like hiring activity or funding stage describe what a company is DOING right now, a better proxy for timing outreach.
A tool that shows the cost of a query before running it prevents a common lead-gen failure: paying for a broad search that returns mostly unusable results.
12Unipile
Reusing an existing personal account via QR-code login sidesteps a lengthy platform approval process, but the account's own rate limits and policies still apply in full.
Outbound automation on a platform with a stated daily cap needs that number verified directly with the platform, not taken on secondhand authority, since limits change and enforcement varies.
Glossary
Terms worth knowing.
MCP
Model Context Protocol: a connector that gives an AI agent live, structured tool access to an external service's API, loaded into context at the start of every chat.
CLI
Command-line interface: a lightweight command an agent runs directly against a service, avoiding the token overhead of loading a full MCP's API schema into every chat.
Claude Cowork
Anthropic's web-based collaborative workspace version of Claude, distinct from the terminal-based Claude Code environment.
CLAUDE.md
The project instructions file Claude Code reads automatically at the start of every session, making its size a recurring, per-session cost.
BDR
Business Development Representative: a sales role responsible for outbound prospecting, used in the video as an example of a hiring signal worth targeting.
“With the Gmail one you can see we only have read-only tools, we can't actually send emails with this Gmail connector.”
names the exact limitation the paid CLI fixes→ TikTok hook↗ Tweet quote
12:20
“If you do 30 sessions a day that means 900 a month, and then optimizing your CLAUDE.md is going to make a significant difference in your token spend.”
concrete math stat, no hedging→ newsletter pull-quote↗ Tweet quote
13:01
“The caveman shaved 18% of the skill's input tokens with no meaningful quality regression.”
specific proof-point stat→ IG reel cold open↗ Tweet quote
14:25
“I ran a quick test on five Ecom websites and you can see that the web fetch tool from Cloud failed on all while Firecrawl was able to scrape four out of five.”
head-to-head comparison with a hard number→ TikTok hook↗ Tweet quote
16:58
“This cloud video plugin can basically start watching videos, because besides just scraping a transcript it can actually see the video.”
clear, one-line value proposition→ 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
metaphoranalogy
If you're using Cloud Code or Cowork, you have access to an extensive ecosystem of skills, plugins, MCPs, connectors, and CLIs to instantly upgrade its capabilities. But with so many out there, it can be hard to know which ones to actually use. So in this video, I'll break down my top 12 that you might want to use yourself, show you why they matter and what they can be used for, and exactly how to install them for yourself.
Now, the first one is the one that I probably use most, which is the Google Workspace CLI. And this allows you to basically connect every Google product instantly to... Cloud Code with one CLI, which if you're still unfamiliar with CLIs, I'll explain in a second.
We can connect it to Drive, Gmail, Calendar, Sheets, Docs, Chat, and every other Google Workspace API. Now you might be wondering, don't we already have the Google Connector like Gmail, Google Calendar, and the Google MCP available to using Cloud Code Work or Cloud Code? And yes, but there are two big advantages with the CLI.
over these connectors or mcps firstly almost all of these google mcps like gmail have a lot of limitations for example with the gmail one you can see we only have read only tools we can't actually send emails with this gmail connector we also can't actually create or edit any google sheets any google docs you can't create a calendar event for example and there are many other limitations to the google mcp and connectors but with this cli we have none of those limitations here for example i asked claude to use the Google Workspace CLI to check my inbox and make a Google Sheet with all of the sponsorship opportunities from the last 90 days.
And you can see it created a Google Sheet here with all the opportunities and the relevant columns, which of course wasn't possible with the Google MCP because we can't actually create a Google Sheet. But the second advantage of this is that a CLI is also far more token efficient than an MCP. Now, if you don't know what a CLI is yet, it's just another way to connect your software to an AI agent on Cloud Cowork or Cloud Code.
Both an MCP and a CLI. lay on top of an api layer of a software the difference is that with mcps and connectors for example in co -work in every new chat you start it automatically loads the mcp context with all the information on all the apis directly in the context window and this means you spend far more tokens and of course this can also lead to worse outputs while cli doesn't load instantly at every new chat now one thing you want to keep in mind is that these clis work best in cloud code so if you're used to working on co -work easiest is just to switch to the code tab and use the cli in there now how do you install it google has a full guide here on installing it but the installation can be a bit tedious because we actually have to set up a google app with zero auth so i actually created a quick skill that you can download for free in the first link in the description that helps you set it up you can just run the skill you'll then use your browser and do almost the entire process for you all you'd have to do is log into your google account follow the steps let it do its job and then you'll have it set up in a couple of minutes once you have it set up it'll be there forever and you can take any action across any google product anytime you want also
If you're interested in my full library of skills and plugins across marketing, sales, operations, and customer support, together with full cloud courses, unlimited one -on -one live tech help, and a community of professionals and business owners, you can also check out my AI Accelerator in the second link in the description below if you want to dive a little bit deeper into these tools.
Again, you can download this skill and all the CLIs and MCPs that I mentioned together with the links for free in the first link in the description below. Now, the next one is a game changer if you're designing in Cloud Cowork or Cloud Code, which is Higgs Field. fields we basically give cloud access to the best image models like nano banana or gpt image 2 but also to video models like cdance which we can then use right inside cloud they have both an mcp and a cli the mcp we can use in cloud co -work the cli again in cloud code here for example i used it to generate some of my community designs Because every time we upload a new skill, for example, we have to make these designs.
And because my team doesn't really know how to use Figma, they now use Higgs field to generate these. So all I said is use the Higgs field MCP after installing it, of course, with the latest image model, generate five post cover banners, one per title with the five skill images that I wanted to create. And I gave it my brand guide and I fed it one reference image, as you can see here.
And then it instantly generated all of these designs in the right format. And the reason I really like Higgs field and I definitely recommend checking this out if you have frequent design use cases or needs is that firstly, these images are directly displayed in the chat. Second, we can use these reference images, which is actually takes into account to make these designs far more consistent and reliable.
Thirdly, we can also really easily edit or even animate and download the images with these buttons here. So if I click edit here, I can just give a prompt quickly and we can iterate on it fast. And lastly, which is probably the biggest reason.
Because we can use it right inside Cloud, we can actually build skills on top of it to automate repetitive design workflows or needs. For example, here I build a skill for my YouTube thumbnail designs. And whenever I use the skill, it already knows what to ask me.
So I can throw in reference images, for example, because often my thumbnails follow a similar format. So I just throw in a format of an old thumbnail that I'm looking for. I can also throw in specific elements like specific logos or photos of me, which you can directly use in that thumbnail.
Then I just... in text what I want or need out of this thumbnail and how many variations I want. And it generates me some thumbnail variations, in this case with Nano Banana Pro, that I can instantly edit or download.
And because I can put any repetitive element inside of the scale, like for example, my photo or this scale file, if I don't just say in the prompt, use my photo or use the scale file, you can easily pull that from the scale and put it inside a thumbnail. So because you can build these skills around it, you can really speed up repetitive workflows.
But Higgs field can also be used for video, audio. We can animate for websites. We can create slideshows, whatever your needs are.
You can just install in Cowork very simply by installing the MCP. By copying this link, going to the customize tab, going into the connectors and add custom connector. Here then you add the remote server URL and the name.
And once you add it, all you have to do is log in and it will be connected. And for Cloud Code, you can just install the CLI with this command line here. You can start with the $15 plan, which allows you 400 Nano Banana Pro image generations or 23 video generations.
So that brings me to the next one, the printing press. So now that you know what a CLI is, you understand the value of this one because the printing press plugin does three things. First, it has a library.
of 50 plus pre -built clis for softwares that have no public api yet for example we instantly get access to a cli on 11 labs open router digital ocean ebay gumroad shopify and many others but second besides these pre -built clis the printing press allows us to build our own clis really fast and this means any mcp or any software you often use in cloud you can now build a cli out of to save a lot of tokens and to make it more efficient than an mcp But thirdly, maybe even more powerful, the printing press can even be used to build CLIs out of softwares that do not have APIs yet, which of course can be extremely powerful if you're maybe doing that right now through the Cloud Chrome browser, because the Cloud Chrome browser, of course, is far less efficient than using a CLI.
The CLI is more reliable and less token heavy and less error prone than browser computer use. So how do we set it up? Again, we want to be in the code tab and all we have to do is just tell Cloud to set up the printing press from scratch.
I'll also share the prompt that I gave it here to install the printing press in the free resources link so you can just copy and paste this first part and you'll have access to the library of CLIs that they list and then you can also build your own CLIs. So in my case, I instantly instructed it to build a CLI out of my YouTube studio, which is a very complex API and MCP to set up.
That's why I wanted to try and build a CLI out of. And after a few minutes, you can see it installed everything. Here's what now is on the machine, the factory binary, the library with all the CLIs from the printing press, the YouTube CLI that it created.
It also comes with some skills right away that teach basically cloud how to use all of these CLIs efficiently. So again, all you really have to do is paste in that first part of the prompt to set up the printing press, and then you can just ask it to build a CLI out of a software and walk you through the process. Now, the next one is Impeccable, another design use case, which is basically a set of skills for cloud code that help with HTML and website design.
and this one's very useful if you're often building or deploying html based website maybe presentations or one -pagers with cloud code because it basically gives you an easy way to improve any designs that cloud has generated locally by being able to select any element on the html page and then improving it easily here with the most common use cases for design for example we can adjust the layout easily by clicking the layout we can adapt we can polish we can animate and each of these use cases has a skill with best practices already baked in so for example i can do animate i can refine further and adjust any specific things or i can just go and say go and it will then generate three variants and then we can decide what we want to change it then generates the three variations as you can see here one goes from bottom to top one word by word and one zooms a little bit then you can just accept it or tune it You also have a feature where you can add your brand guidelines or your design MD directly in here and then click, for example, detect.
And then it will automatically go through and look for things that can improve based on your design MD. But it can do various things like adapting copy according to UX writing, best practices and other things. So very useful if you're often designing with HTML.
You can install it really easily in PlotCode by just clicking here on get started. And you get three options. You can just copy and paste the command line here.
You can add the plugin or download the zip. If you give the command line to Cloud Code, it will install all of the skills. And then whenever you have an HTML or a local host, you just tell it to run the impeccable skill.
And then you'll get that visual overlay where you can select the elements. Which brings me to the next one, which is relevant probably if the previous one was relevant for you too, which is Vercel. Now Vercel basically allows you to turn any HTML you've generated into a live URL in around 10 seconds by deploying it onto an actual server.
So basically anything that you're building in Cloud Code that you want to host and distribute externally, whether those are reports, websites like these, proposals, dashboards, one -pagers or onboarding guides, you can deploy them instantly onto a server and a URL. through Verso. And I highly recommend starting to do this for things that you want to distribute externally, because of course we can do this through PDFs or slideshows, but they are static and are pretty limited in terms of what we can do to the design with them.
And they can also not be live updated. And with this setup, of course, HTML is the native language for these LNM models. And second, once we have it deployed on the server, we can also instantly update it and have a lot more flexibility in what we want to do with that design.
for example we use this for client onboarding pages dashboards and personalized sales proposals So here, for example, I have a skill that goes through the meeting transcript of a past client, then generates a branded HTML, and then deploys it instantly on Verso, which is instructed in the skill. And then it gives me this public URL link, which I can then actually share with the client.
Again, it's through a CLI on Cloud Code, really easy to install. You can just copy and paste this command line into Cloud Code. Again, the link for this install will be in the free resources.
Then we have the Caveman plugin, which is a set of cost -saving skills, which you can give to Cloud. It basically compresses any input text by what they say up to 75%, but in my experience, it's a bit less, by basically talking like a caveman and taking out filler words while still keeping the same meaning. So you can, first of all, apply this to how Cloud responds to you to save some tokens there.
You can just say, talk like a caveman after you install the skills, and it will then talk like a caveman, and you might save some tokens. But I'm not a huge fan here because it's not very readable. so the way you should be using this in my opinion is to use it on input text because that's where the majority of your token spend goes and second for input text that you frequently use basically to compress them to make sure that in these frequent use cases it'll spend less tokens so the first one that is a good use case for the caveman skill is the cloud md of course because if you're using a folder this is of course the file cloud reads every single time and applying caveman to the cloud md is going to save you money over and tokens over the long term because let's say if you do 30 sessions a day that means 900 a month and then optimizing your cloud md is going to make a significant difference in your token spend so here for example i ran the caveman scale on the cloud md and i now have a 40 percent shorter
Cloud MD. Of course, you have to make sure that it still works well, but usually it does. Second, you can think about applying the caveman to common skills you're using.
Skills can be also token heavy and are usually filled with filler words that are not necessary for still keeping the same functionality. For example, here I ran caveman on a skill and asked it to test this new skill right away versus the original skill to actually make sure the quality doesn't go down of the skill, which is what I recommend you to do to make sure you don't end up with a worse performing skill.
And as you can see, the caveman shaved 18 % of the skills input tokens with no meaningful quality regression. Thirdly, you can also think about any repetitive context file you give Cloud that you often use that is maybe very large. You can run the caveman on.
And another good use case, which they have a specific skill for in the plugin, is to compress MCP documentation, which can also be very token heavy. You can install them really easily in Cloud Code by just copying and pasting this. And then you can also use them in Cloud Code work.
Again, I'll put the link in the free resources in the first link in the description. now the next one you might have already heard of but is a must -have in my opinion which is the fire crawl mcp or cli now fire crawl is a cheap and really good web scraping tool And you might be asking, why do I need it if Cloud can already reach the internet?
It's because Cloud's native web fetch tool actually fails to access a large amount of websites. For example, many e -com websites have JavaScript rendering, which Cloud can't actually access. And second, Cloud's native web fetch tool gets raw HTML, which wastes a lot of tokens.
So Firecraft basically allows you to scrape 99 % of websites, including the ones Cloud can't, and also returns scraped websites in Markdown in a cleaner way. that Cloud can process directly with less token burn. You can use it on a thousand pages per month on the free plan.
So for any web scraping, Firecrawl, in my opinion, should be your default web scraper. So here I ran a quick test on five Ecom websites and you can see that the web fetch tool from Cloud failed on all while Firecrawl was able to scrape four out of five. It also has some other use cases like search, crawl deep research and other functionalities the mcp and the cli are easy to install i'll make sure to link the the page again into free resources but for the mcp and cloud desktop you can just copy this put in your fire crawl api key here once you've set up an account and you just set them up in the custom connector here by adding in the link with your api key and for the cli and cloud code we can just copy this command line now one note firecrawl can't scrape social media so for social media you always want to use appify now i haven't included the appify because i know most people already know about it but appify is the go -to place to scrape any social media page and again has an mcp and i think a cli available too so i believe appify and firecrawl are the main scraping tools i believe anyone needs to have set up now the next one is the playwright cli which is a great alternative for anyone who's using
the cloud browser a lot and the playwright cli is basically a browser automation library but in contrast to cloud's native browser use feature it actually takes actions through python scripts and it can do anything on any website even websites without apis from scraping to form filling to clicking buttons navigating flows and taking screenshots and because it takes actions through python scripts it's far faster less error prone and cost efficient than the cloud browser use so for websites that you're constantly accessing through the cloud browser you probably should be using a playwright cli here for example i use it to go through five local businesses and fill out their websites forms to do lead generation which can be an interesting strategy of course because most businesses actually read those forms that are filled out because for them usually it means leads of course i also gave it the playwright cli it then installed it and then went through all five websites and filled out their forms as you can see here
Now, this is just one use case for Playwright, but it's especially powerful for repeatable scraping or action workflows because it actually writes those Python scripts, which it then can reuse each time on that website. For example, if you often have to scrape large amounts of data on e -com websites and use cases like that.
Again, you can easily install the CLI with the command line in Cloud Code. They also have an MCP available, which you can find the documentation here. It's an open source tool and completely free to use too.
Now, the next one is another scraping plugin. I've covered many scraping tools, but none of them can actually scrape videos, which is what this one solves. this cloud video plugin can basically start watching videos because besides just scraping a transcript it can actually see the video by downloading the video and then taking screenshots so it has full context not just a transcript for example here i asked claude to create a competitor intelligence google sheet that scrapes instagram reels from some ai creators in my niche And then for each analyze the hook and why it worked together with a quick screenshot in the sheet.
So I understand what visually worked. The plugin then bundles a set of skills and CLIs to do this, for example, to transcribe the video and to get the screenshots. And then it gave me a Google sheet here with the full analysis of the reels together with the hook frame.
So I can actually see what kind of visual was shown. And because Cloud can actually see the video with these screenshots, it can then give me a summary on why this reel actually worked. Now, especially if combined with maybe a platform like Appify, this can be really useful for scraping social media videos or analyzing videos on different platforms.
Again, it's a plugin with skills, mostly for Cloud Code. and can be added really easily by just copying this. Again, the link you'll find in the free resource.
Now, one more scraping tool I recommend for anyone doing lead generation or outreach is Vibe Prospecting. It can be used for building lead lists, finding contact info like emails, enriching leads, can be used for recruiting or personalized outreach. And Vibe Prospecting is basically a sales or lead database sort of like Apollo, but it's natively built for AI agents like Cloud so we can query in natural language.
But also while most of these lead databases like Apollo allow you to... filter based on basic criteria like industry, company size, geography, etc. Vibe prospecting actually allows you to filter with up -to -date intent signals.
For example, filtering on current hiring trends, recent funding rounds, and more. For example, here I used Vibe prospecting to find five US SaaS companies with 10 to 50 employees that are hiring BDRs right now. And you can see I found five leads that are hiring for actual BDRs right now.
So for any repetitive enriching or lead gen, use case i think can be very interesting especially if combined maybe with skills routines or scheduled tasks now you'll pay according to the credits used depends a little bit on which use case you use it for how many credits you're using but the tool will almost always show you first before the query before actually running it how much credits it's going to use but definitely worth checking out if you have lead gen enrichment or sales use cases you can install it really easily in the cloud desktop in either code or cloud co -work or just going to the connectors because they have a native connector set up so we can just add the vibe prospecting here now the next one is unipow and unipow allows you to connect cloud to whatsapp instagram linkedin and some other platforms to either read your messages but more importantly it can actually send messages now the big advantage with unipow especially for whatsapp is that it works with the unofficial whatsapp api so you can use your current number and connect it with a simple qr code just like you would do with whatsapp web so you don't have to go through the tedious
official WhatsApp business API process. And for Instagram and LinkedIn, it can do active outreach and send DMs from your account. Now don't actually use this internally because we don't do any outreach right now on LinkedIn, but we often use this for WhatsApp and LinkedIn use cases for our agency clients.
For example, I could say something like send a message to my co -founder, Oscar Johnston through LinkedIn with the Unipal MCP. This is a test for my plugin video.
You can see it now sent a message to my co -founder here. So especially, I think again, if you combine this with maybe scheduled tasks or routines in cloud code, you can come up and automate some of your LinkedIn outreach or your WhatsApp customer support or sales support. It will cost you $55 a month, but if you have this WhatsApp use case or LinkedIn use case, in my experience, this is the best.
tool or mcp to use and you do have to be careful with linkedin outreach of course because linkedin has restrictions and limitations which i believe is 20 a day but please recheck it before trusting my word you can set it up by just creating an account and then you can go to the resources dev tools and there you'll find the mcp and again i'll make sure to put the link also in the free resources now if you want access to all of the skills and plugins that me and my team are building out across all business departments together with one -on -one unlimited live tech help courses multiple weekly q and a's and a community of professionals and business owners you can also check out my ai accelerator in the second link in the description below thank you so much for watching if you got any value out of it i highly appreciate a like and subscribe and if you want to learn more about working with cloud you can check out the video here above
The Hook
The bait, then the rug-pull.
A growing ecosystem of skills, plugins, MCPs, and CLIs can upgrade what Claude Code and Claude Cowork are capable of, but most of it is noise. This breakdown walks through twelve specific tools, what each one unlocks over the default connectors, and exactly how they get installed.
Frameworks
Named ideas worth stealing.
06:22list
Printing Press's three functions
Pre-built CLI library for 50+ tools with no public API (ElevenLabs, OpenRouter, DigitalOcean, eBay, Gumroad, Shopify, and others)
Build a custom CLI for any MCP or software already used often, to save tokens versus the MCP
Build a CLI out of software that has no public API at all, replacing slower browser-based computer use
One plugin covering three distinct capability tiers: use what's already built, build your own from a known API, or build one from a tool with no API at all.
Steal fordeciding whether a frequently-used integration should be installed, custom-built, or wrapped from scratch
01:50concept
MCP vs. CLI tradeoff
Both an MCP/connector and a CLI sit on top of the same underlying API. The difference is when the integration's context loads: an MCP loads its full schema into every new chat automatically, while a CLI only loads when actually invoked, which costs fewer tokens and avoids degrading output quality with unused context.
Steal fordeciding whether to add a new integration as an always-on connector or an on-demand CLI
CTA Breakdown
How they asked for the click.
VERBAL ASK
21:16product
“if you want access to all of the skills and plugins that me and my team are building out... you can also check out my AI Accelerator in the second link in the description below”
soft pitch placed after all 12 tools were delivered in full, paired with a like/subscribe ask rather than fronting the offer
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.
A 43-minute numbered walkthrough of all 34 Claude Cowork concepts across memory, automation, connectors, and team rollout, framed as a business operating system not a chatbot.
Why showing Opus 5.5 a finished example now beats writing out the steps, proven by rebuilding a newsletter skill from six best and six worst performing emails.
A YouTuber pulls Anthropic's own internal prompting guides for Opus 5 and Fable 5, cross-references them against Claude Code creator Boris Cherny's Y Combinator talk, and distills them into seven concrete habits.
A 32-minute full guide to building a Claude-powered second brain: folder structure, daily auto-updates, vault hygiene, team sync, and autonomous cloud routines.
00:10
00:16
00:39
00:55
01:17
01:39
02:01
02:54
03:07
03:18
03:40
04:22
04:54
05:16
05:36
05:49
06:10
06:51
07:08
07:25
07:41
07:58
08:32
08:49
09:07
09:24
09:41
09:58
10:31
10:48
11:04
11:20
11:36
11:52
12:08
12:55
13:11
13:27
13:42
14:12
14:28
14:43
14:54
15:20
15:28
15:43
15:54
16:28
16:41
16:58
17:13
17:44
18:03
18:15
18:51
19:08
19:26
19:58
20:14
20:30
20:48
21:01
21:10
43:17
09:23
13:15
32:06
13:39
24:27