The argument in one line.
After learning these 20 concepts in order from workspace folders to sub-agents, every Claude Cowork tutorial will finally make sense because you'll understand the foundational building blocks that all advanced techniques rely on.
Read if. Skip if.
READ IF YOU ARE…
- A beginner exploring Claude desktop for the first time who feels lost by tutorials that assume you know what workspaces, skills, and sub-agents already mean.
- Someone building AI workflows who wants a foundational mental model of all 20 Cowork concepts before diving into advanced implementation.
- A non-technical person or small business owner considering Claude Cowork but intimidated by the jargon and feature density.
SKIP IF…
- You already use Claude Cowork regularly and are looking for advanced patterns, optimization strategies, or edge-case troubleshooting.
- You need instruction on connecting Cowork to external APIs or databases — this is a concept glossary, not a systems integration guide.
The full version, fast.
Claude Cowork's twenty core concepts stack into a coherent system once you stop treating them as jargon and walk them from simplest to most powerful. Start with the workspace folder Claude reads and writes into, then layer a project-scoped Claude.md and account-wide global instructions to define behavior, memory to persist what matters across sessions, and the context window as the desk where it all sits. Build outward through multimodal inputs, web search, extended thinking, and artifacts that produce real dashboards instead of paragraphs. The leverage compounds at the top: projects isolate brains, skills encode repeatable workflows triggered by slash commands, plugins bundle skills, connectors wire in your actual apps, Chrome and computer use cover whatever lacks an API, scheduled tasks replace Zapier, dispatch controls everything from your phone, and sub-agents run those skills in
Free for members
Chat with this breakdown — free.
Sign in and you get 23 free chat messages on us — ask for the hook, quote a framework, find the exact transcript moment, generate a markdown action plan. Bring your own key when you want unlimited.
Create a free account →Where the time goes.
00:00 – 01:07
01 · Cold open + promise
Names the overwhelm (MCPs, skills, sub-agents) and promises 20 concepts simple-to-advanced.
01:07 – 03:07
02 · #1 Workspace Folder
Where Claude's files live on your computer; create a folder, point Claude at it.
03:07 – 05:08
03 · #2 Claude.md
A markdown file that defines how Claude operates; loaded at the start of every session.
05:08 – 06:34
04 · #3 Global Instructions
Permanent identity across all Claude surfaces, not project-scoped like Claude.md.
06:34 – 07:47
05 · #4 Memory
Markdown files Claude builds over time; tell it 'save that to memory' to keep context.
07:47 – 08:34
06 · #5 Context Window
1M tokens — the 'desk' for each conversation. When full, older context drops.
08:34 – 09:09
07 · #6 Multimodal
What Claude can see: images, PDFs, text, screenshots. Not video yet.
09:09 – 09:42
08 · #7 Web Search
Live web lookup since Claude's knowledge cutoff is May 2025.
09:42 – 10:09
09 · #8 Extended Thinking
Claude takes longer on hard questions; better answers at the cost of latency.
10:09 – 11:32
10 · #9 Artifacts
Non-text outputs (HTML dashboards, slides, charts) rendered inside Claude.
11:32 – 13:46
11 · #10 Projects
Separate brains in one workspace — each project has its own memory, instructions, scheduled tasks, files.
13:46 – 14:22
12 · #11 Bash Tool
How Claude runs code itself instead of handing you commands to copy.
14:22 – 16:48
13 · #12 Skills
One command that triggers a whole workflow — Brock demos his morning briefing skill.
16:48 – 17:17
14 · #13 Slash Commands
Keywords that trigger skills — e.g. /morning-briefing.
17:17 – 18:52
15 · #14 Plugins
Bundles of skills you can share or upload as a zip — easier than re-creating 50 skills.
18:52 – 23:42
16 · #15 Connectors
Hooks Claude into your apps (Gmail, Calendar, Slack, etc); MCP servers like Zapier MCP expand the catalog.
23:42 – 25:10
17 · #16 Chrome Extension
Claude controls your browser to click, fill forms, book things — one step below connectors.
25:10 – 26:55
18 · #17 Computer Use
Claude takes over your whole computer (not just browser) — drag files, use apps. Slow but powerful.
26:55 – 29:00
19 · #18 Scheduled Tasks
Cron-like automations built on skills — Brock's daily wrap-up auto-runs every evening.
29:00 – 29:59
20 · #19 Dispatch Mode
Control your desktop Claude from your phone — trigger skills on the go.
29:59 – 30:47
21 · #20 Sub-Agents
Parallel workers for complex tasks — research / financials / narrative / slides in parallel.
30:47 – 31:27
22 · Outro + CTA
Subscribe + join School community for 50+ weekly-updated skills.
Lines worth screenshotting.
- Claude Cowork's 20 concepts span from workspace folders to sub-agents, and understanding them in order from simplest to most powerful is the fastest path to fluency.
- A CLAUDE.md file is the project brain — loaded at the start of every session, it defines how the AI operates within a specific workspace.
- Global instructions differ from CLAUDE.md in scope: they apply across all of Claude (chat, Cowork, Code), while CLAUDE.md is folder-specific.
- Selecting multiple folders inside Cowork expands the AI's file access as complexity grows, rather than requiring a rebuild of the workspace.
- The most common source of overwhelm in Claude Cowork is learning concepts out of order, making earlier ideas seem more complex than they are.
- Sub-agents are the most powerful Cowork feature because they enable parallel task execution, but they are only useful after the foundational concepts are internalized.
- Skills inside Cowork are reusable instruction sets that tell the AI how to perform a specific task consistently without re-prompting from scratch.
- MCPs (Model Context Protocol connectors) extend what Cowork can access — web browsers, APIs, external tools — turning a text interface into an action-taking agent.
- A beginner who understands the workspace, CLAUDE.md, and global instructions can extract 80% of the value from Cowork before ever touching skills or MCPs.
- Connectors inside Claude Cowork translate between the sandbox and external services, which is why the Higgsfield connector was such a significant release.
- Treating each Claude project as a separate AI employee with its own CLAUDE.md and skills prevents context bleed between unrelated workflows.
- A numbered glossary format is the right pedagogy for Cowork onboarding because it gives beginners a clear map of what they do not yet know.
Steal the numbered-glossary format.
Brock's beginner-glossary playbook
Pick a tool ecosystem with too many confusing words, number all 20, walk simplest-to-advanced, branded slide for each — own the SEO term forever.
- Pick a tool ecosystem where beginners feel lost (Supabase, Trigger.dev, $6 Stack, MCN+, Claude Code itself) and list every confusing concept on a whiteboard.
- Sort 1-20 from absolute basics to most powerful — never re-order, never skip — the numbering itself is the promise.
- Build ONE branded slide template ('CONCEPT 02 — TITLE — one-line tagline') and reuse it 20 times. The repetition is the brand.
- Pair every concept slide with a 30-90 second live demo inside the actual product UI. Glossary alone is content; glossary plus demo is a tutorial.
- Repeat one comfort-line across the video ('I was in your exact same shoes') as a re-hook between concepts to reset attention.
- End on the most-mind-blowing feature (sub-agents) — never bury the wow at concept #7.
- Soft-CTA your paid offer 4+ times by weaving it into demos ('this skill is in my school community'). No hard pitch block needed.
- Steal candidates: 'Every $6 Stack concept', 'Every MCN concept for creators', 'Every Trigger.dev concept', 'Every JoeFlow concept'.
Terms worth knowing.
- Claude Cowork
- A GUI-based agentic feature inside the Claude desktop app that lets non-developers give Claude access to files, the web, and external tools to complete multi-step tasks autonomously.
- workspace folder
- A folder on your computer that you connect to Claude Cowork as the root directory — all files Claude reads, writes, and creates during sessions live here.
- Claude skill
- A Markdown file containing packaged instructions for a specific task that Claude Cowork can invoke automatically when a relevant request is made, without re-prompting each time.
- MCP (Model Context Protocol)
- An open standard that lets Claude connect to external data sources and tools — such as databases, APIs, and services — through a unified interface, extending what tasks it can perform.
- connector
- A pre-built MCP integration that links Claude to a specific external service (like Google Drive, Notion, or Slack), allowing Claude to read and act on data from that service.
- sub-agent
- A separate Claude instance that a main Claude session can spawn to handle a specific sub-task in parallel, with results returned to the orchestrating session when complete.
- AI employee
- Informal term for a configured Claude Cowork session or agent set up to handle a specific recurring business task autonomously — a delegation metaphor rather than a literal product feature.
- CLAUDE.md
- A Markdown configuration file in the workspace folder that acts as persistent instructions Claude reads at the start of every session, defining its role, rules, and available resources.
Things they pointed at.
Lines you could clip.
00:27
“Skills, MCPs, sub agents, connectors — everyone throws out these words like you're supposed to already know what they mean.”
00:30
“Every tutorial you watch from here on out will finally click.”
13:46
“One command that can trigger entire workflows for you.”
29:00
“I've replaced n8n and different automation platforms with this.”
30:47
“Parallel workers that are working at the same time.”
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.
metaphorstory
00:00Cloud Cowork is one of the most powerful AI tools out there, but if you're new to it, it can feel extremely overwhelming. You've probably watched a bunch of different tutorials on it, and half the time, you have no idea what it is they're talking about. Skills, MCPs, sub agents, connectors.
00:15Everyone throws out these words like you're supposed to already know what they mean. So in this video, I'm gonna break down all 20 Claude co work concepts, starting from the absolute simplest and building all the way up to the most powerful features.
00:27By the end of this video, you'll actually understand what co work can do, and every tutorial you watch from here on out will finally click. So without further ado, let's get in. Alright, guys.
00:37So I just wanna start off and say I know exactly how you are probably feeling. There's tons of different concepts inside of Claude and specifically Claude Cowork, and it can be really overwhelming to understand. And I was in your exact same shoes, so this video is really gonna simplify it so you understand to the core all these different concepts so you could begin using this and implementing it into your Claude Cowork and begin building different AI employees inside of this platform.
00:59So let's jump right in and talk about the first concept that we need to understand, and this is where you start when you begin using Claude Cowork. So let's just talk quickly about the workspace folder.
01:11So this is basically where all of our files are going to exist inside of Claude. The cool thing about Claude Cowork is that you can access actual folders and files on your computer. So for example, like, you could see right here, this Claude code short system folder.
01:26This is where every single file that I generate inside of Cowork lives. So we could pull information from here. We could add different information, all those different things.
01:35This is basically where it lives. Inside of these things can be document files, HTML files, Python scripts, CSV files. Basically, anything that you can create on your computer, it lives inside of this.
01:46So first things first, we need to make sure we have the Claude desktop app downloaded onto our computer. There's gonna be a link in the description to download it. If you don't already have it, you would just click download right here.
01:57You could download it for Mac. You could download it for Windows. All those different things, just make sure to download it, and then I'm gonna pull mine up.
02:03We're gonna see normal chat mode up here on the top. This is basically where we chat back and forth with Claude as if we're talking to Chatuchupti, but we're gonna focus on coworkers. So make sure to toggle this on, and you could see this interface here.
02:16And the key thing I wanna mention is this is the workspace folder right here. So if we don't already have any folders, we could actually choose or create a specific folder. So these right here are basically all of the different folders on my desktop, on my MacBook.
02:31So we could select a preexisting one if we want to. However, if you're getting started, I highly suggest just creating a new folder here. Let's just call this Claude Cowork folder.
02:40I'm gonna click on create. You could see that it's actually empty here. There's nothing living inside of this.
02:45As we begin using this more and more, different files will populate inside of here. I'm gonna click on open. I'm gonna click allow.
02:52And just like that, we have our Claude CoWork folder open. And the key thing here, we could actually select multiple different folders if we wanna work inside of them.
03:01So as you expand, you could add different folders. But for right now, just focus on that one folder we created. Moving on to concept number two, and you've probably heard this tons in any of the Claude CoWork videos that you have watched, and that is Claude MD.
03:16And this sounds pretty complex, and let me tell you, it's super simple. Once you get the grasp of it, it's gonna make so much sense and make your experience so much easier inside of Claude. So this is basically a file that defines exactly how Claude works, and this is loaded at the beginning of every single session.
03:33So think of a prompt as what you send to Claude or any large language model like ChatGPT, for example. This is basically just a document that's inside of the folder that you just created that teaches Claude exactly how you want it to operate.
03:49It will then read this every session before every message that you send, and this is how you have a personalized Claude experience based on who you are, how you want it to interact, and all those different things that you need to specify. So for example, this is what it would look like. You would have your CloudMD that's loaded every session.
04:05It would then have something like who I am. For me, it says YouTube creator, AI and tech content, nontechnical audience. And then it also has my rules here, basically saying, always do this.
04:15Never do this. And, again, this is loaded at the beginning of every single conversation. So if I come back over to co work, let me show you exactly what I mean.
04:24So if I come over to a preexisting task or conversation on the left hand side of our co work interface and click on this, you could see that I already went back and forth in this conversation. But what I could do is I could click on this instructions, and you could see right here it says Claude dot m d.
04:38And this right here is basically that Claude MD file, which is just a set of instructions on how I want it to operate. So this specific CloudMD file is for school. Basically, breaks down what my school community is, a different pricing, revenue goal, breaks down what my members get, basically, everything it needs to know in order to understand my school community.
04:56And this is loaded every single time I have a conversation in this different project. Alright. So we talked about two different concepts.
05:02Hopefully, you are not overwhelmed. You could see I'm gonna very simply break down these concepts, and then you could begin building on top of these. So next up for concept three, we have global instructions.
05:12This is kind of similar to the clotMD file that we talked about. So, basically, it's the same thing in regards to, like, you set it up once, and it references this for every chat conversation that you have. But we need to quickly talk about the distinction between clotMD and a global instruction because they are a little bit different.
05:28So CloudMD lives inside of that specific project folder, and it only applies to that project. It has different rules per workspace that it's inside of, and this is co work only.
05:40So think of your Claude MD as the . And then your global instructions is your permanent identity across all of Claude, whether that is the Claude chat mode, Claude co work, or Claude code. It's basically just global instructions on the entire system.
05:54To show you where you will find your global instructions and how to set them up, you're gonna come down here, select on your plan. You're gonna click on settings. You're then gonna come down and select co work.
06:05And right here, we are going to see our global instructions. You could see that it says instructions here apply to all Claude co work sessions. Use this for preferences, conventions, or context that Claude should always know.
06:17And then we could come over here, click on edit, and basically just add in all the context we wanted to have. I would just say, like, who you are, what you want Claude to do, how you want it to actually respond to you, if you wanted to be more friendly or you wanted to sound more professional. These are all things that you could load inside of your global instructions.
06:34Next up, we have memory for concept number four, and this is basically how Claude understands all the conversations you have with it, has context on different things that you wanted to do and you have done previously, all these different things. This one's not rocket science. Memory is exactly what it sounds like.
06:51So if we head back over to Cowork, I'm inside of this project, and we're gonna talk about what projects are later on in this video here shortly. But if I come over to the right hand side, you could see we have this memory section. And this is basically a bunch of different markdown files with memory on how I like some different things to be done.
07:09So whether this is, like, context on yourself or maybe a specific workflow or task that you create, basically, you could just say, hey. I want you to add that to my memory. And memory is something that will just build over time.
07:19If there's something you want Claude to remember, I'd just say save that to memory to make sure that it does save it, but it should save the important things over time. You could see right here, this is, like, a very niche thing. But feedback on a slide generation skill that I had when it said don't use these phases ever, how to apply all these things.
07:35This is just memory. You could see this built up inside of projects, which we're gonna talk about later in this video. So this is basically how Cloud remembers everything and all the different conversations you have with it.
07:45Alright. So moving on to concept number five. This is the context window.
07:49I want you to think of the context window that you have with Claude. You can only fit so much on that desk. And right now, you could fit up to 1,000,000 tokens, which is basically what it's called.
08:02And things that are inside of the context window are things like the Claude MD, memory files, system prompts, any conversation that is coming both in and out of Claude, different files and documents that we have in there, and different tool outputs that we use with inside of our conversation. So right now, have a 1,000,000 token window.
08:20This will probably go up over time, but you will probably run into issues if you're going back and forth with Claude for quite a long time where it will need to condense that context window, and we'll have to, like, push out certain irrelevant information in order to, like, refresh it. Moving on to concept number six, and this is multimodal.
08:37If you've been around AI for a little while, you've probably heard this. For me, this was confusing at first. It sounded super complex, but it's really not.
08:43It's as simple as all the things that large language models and specifically Claude can see. So think of things like images, screenshots, charts, PDFs, different text you give it, all those different things.
08:55And video would be another option, but that's actually not an input format that Claude is actually able to see. So when people say that Claude is multimodal, that basically just means that it can read files, it can see text on screen, and it could even see different images that you upload. Next, moving on to concept number seven.
09:11This one's pretty self explanatory, but it's called web search. So this is basically how Claude is able to search the web for different live information.
09:19So as of right now, the knowledge cutoff for Claude is May 2025, so it doesn't know about different new AI tool releases or different current pricing for specific models, different things like that. So it's actually having to use the web search tool, and that's where it's able to go off and scrape and find different information from the Internet, pull different articles, all those different things.
09:39It's really simple. That is the web search concept. Now moving on to concept number eight, we have extended thinking, and this is where Claude can think through more complex problems and do more complex tasks for you.
09:51This is as simple as asking a hard question. Claude will think very deeply about it, go back and forth, ruminate about it, take a bit longer, but it will come out with a better answer. Think of it like asking somebody to take their time.
10:03They're just gonna get the task done better, but it will obviously take more time for them to give a response or to do something. Moving on to concept number nine, and I absolutely love this concept. It is artifacts.
10:15So an artifact is basically an output inside of Claude that is not just like text. If you've ever chatted back and forth with ChatGPT, it'll just respond with paragraphs and paragraphs of text, whereas an artifact is an output such as, like, an HTML dashboard or a chart or visual, something like that that you could actually see inside of Claude.
10:33And if I come back to co work, you could see that I have this conversation here. And if I pull this up, this is an HTML dash board breakdown of my YouTube competitors and different videos that people are making throughout the day.
10:46So that way, I could see what is hot in the AI space, what should I pay attention to, and try to learn so I could give back to my school community. This right here is an artifact. It's an HTML dashboard that it generates.
10:55And if you're new to Claude specifically, but even more so Claude Cowork, the fact that we have artifacts baked into all the conversations we have is just so amazing. You're gonna love it. And the amount of the different artifacts you could create are actually pretty mind blowing.
11:08This slideshow that you guys are seeing actually in this video is an artifact. You could see I created this inside of Claude Cowork, and you could see on the right hand side our artifact here, and we could even open this up in our browser and view it. That is basically what an artifact is.
11:21Alright. So next up, and this is where things get really interesting and exciting because this could refine how we use Claude CoWork. So instead of just using it very simply, these are tips and tricks you could use to be a power user of it.
11:32So our next concept is project. So think of these as separate brains inside of one specific workspace. So each project has its own memory, instructions, and conversations directly inside of co work.
11:45So to show you what I mean, let me come back to my co work here. And if I come over on the left hand side, we are going to see projects in our account.
11:54And I have a couple of different projects here. I have short form. I have school community, Loom agency, and YouTube videos.
11:59And if I pull up my YouTube videos project, there are a couple of different things we could see here. So we have our normal chat window here where we could go back and forth with coworker. We have our outputs, which are basically the artifacts that I just touched on.
12:12So I pulled up. You could see we have our YouTube competitor intelligence artifact that was automatically generated, as well as we have all of our different tasks and conversations inside of our YouTube videos project.
12:23And on the right hand side, we have specific instructions for this project or ecosystem. And then we even have scheduled tasks here that we have running, which I'm gonna talk about later in this video for YouTube video related tasks. And then we even have all of our files inside of this YouTube video project here.
12:39So think of this as a more organized way to use co work. So if you have a specific category that you wanna keep separate from your personal life or your business, I suggest you make a project about it.
12:52You could see that I have a project that is for YouTube videos. I have one that is for my school community. So every single thing school community related is inside of here, all of my scheduled tasks, all of my files.
13:05And I have specific memory on school, so it understands everything it needs to know about my school community. So in order to create a project, you could do a couple of different things. So you could come over to new task and then select the folder selector here, and you could go down and select a project.
13:20You could see we can create a new project as well, or we could just simply click on projects here, come over to a project, begin conversating right here, or we could take an existing conversation that we have, come over here, and then we could also move this to a project as well. And this is just a way to stay much more organized so everything doesn't blend together.
13:39Because if you come over here, all of these chats can get overwhelming, and it's really nice to categorize them and break them up. Next up for concept number 11, we have the bash tool. And this is basically how Claude writes code inside of Cowork.
13:54So instead of Claude giving us code for us to run-in terminal, it will use the bash tool to do this automatically. And, honestly, you don't really need to understand what's going on here, the nuts and the bolts. Just understand this is what the bash tool is.
14:08It's basically how it helps you run code. A perfect example of when this will run is, say, you want to resize different images, rename every file in this folder, scan a bunch of different sales data.
14:19All of these things will then trigger the batch tool to then write the code. Moving on to concept number 12, and this is probably one of the most important ones. This is skills, and this is really what makes Cowork so powerful, much more powerful than any other AI tool out there.
14:33Think of it like this. This is basically one command that can trigger entire workflows for you. So I have something called a morning briefing task that will trigger an entire workflow and then give me a specific output, and I'm gonna show you exactly what I mean right now.
14:47As well as we have a couple other ones like invoice generator, receipt scanner, all these different skills are predetermined workflows that Claude knows how to perform for us.
14:55Alright. So to show you exactly what I mean, I have a skill right here that runs every single morning. So let me break it down very simply for you.
15:02Whenever I just type in morning briefing, this will trigger this workflow where it will go, it will search my Google Calendar. It will then search my email. It will then search all the different AI trending news that I should cover or that I should know about, and then even writes out my top three priorities.
15:18And on top of that, it spits out this HTML dashboard for me, and it even sends me this entire breakdown on Slack every single time that I trigger it. And you're probably asking, like, how does this work? How do we set this up?
15:30So let me show you right now. So if I come over to customize on the left hand side of co work, we are going to see skills right here. And I have a ton of different skills that I have built personally.
15:40So if I come down and find my morning briefing skill, I can click on it, and it will populate this set of instructions. This is basically a markdown file that you could see breaking down everything here. So it says you are a personal chief of staff.
15:53Every morning, you scan the user's calendar. Basically, breaks down what to do. It even breaks down the specific output that I want for that HTML dashboard.
16:01So you could see that this is an Apple Swiss style design. I could change this if I wanted to. But the key thing here is that it will be outputted like this in this style every single day when I trigger it.
16:12So this is just like a a preset of instructions. So that way I could have this workflow run every single time that I trigger it, and it will function exactly the same every time. And if you guys are interested in all the 50 skills that I use every single day inside of my business, there's gonna be a link in the description to join my school community where I update you guys every single week with all the skills that I'm using personally that you guys can access and download.
16:34A couple of the different ones that I have that might interest you are proposal generator, contract review, invoice generator, and tons more across different categories that you guys could begin using and plugging in directly into Cowork and start using right away.
16:48Alright. So concept number 13, and this is kind of building on top of the skills that we just talked about, and these are slash commands. A slash command is basically think of it like a trigger word that will then trigger a specific skill, which is basically that workflow.
17:04So I just give the slash command morning briefing, and it automatically runs that workflow or skill for me that might sound complicated when you hear the word slash commands, but it's as simple as just, like, a keyword to then trigger a workflow. Moving on to concept number 14 inside of Claude Cowork, and that is plug ins.
17:22Basically, what a plug in is, I've had a lot of people ask me, like, Brock, what is the difference between a plug in and a skill? When should I use each of them? And to break it down very simply, plug ins are basically just a bundle of skills together that makes it very easy for you to share with teammates or with a friend, for example, and for you to upload into co work.
17:41So that way, you don't have to individually create 50 skills or upload those 50 skills yourself. So for example, all these 50 skills that I show inside of my school community, you could download this plugin here and automatically upload that into Cowork, and you'll have all of them available.
17:58So in order to do that, what you would do is you just, like, save this zip file here. You would come to co work, and you could see here that we have a bunch of different plugins. We could browse plugins.
18:08These are plugins that Anthropic already has. So we have some for design. We have legal marketing.
18:14All these these are basically just a bundle of different skills for different industries. And if you wanna add your own plugin, you could just come over here, click on personal, click on upload plugin, and then I would just drag in that ZIP file here and upload it. And then all of those 50 skills I just referenced would then be inside of that plugin.
18:31To show you one that I have downloaded, there's this legal one that Anthropic created. And, basically, there are a couple of different skills in here. I could click on all of them, and there's this contract review one that I actually use quite a bit.
18:42And, basically, I could upload a PDF of a specific contract, and it will automatically review it along with redline specific things that I need to negotiate in order to, like, actually sign this contract. Alright. So moving on to concept number 15, and that is connectors.
18:56This is where things really come together, no pun intended. And this is where we just connect all of our applications to coworkers so we can go off and perform tasks for us. As I mentioned in my morning briefing skill that I have, it pulls from my email.
19:09It pulls from my Google Calendar. It posts to Slack. The way it's able to do this is through connectors.
19:14So in order to use connectors, what you're gonna do is you're gonna come over and click on customize. From here, underneath skills, we are going to see this connectors button, and it's very simple to add our different apps. You could see I have a couple of different ones added here.
19:29Basically, I could come over here, click on browse connectors, and we could basically add any of these different apps here. There are plenty that Anthropic actually gives us. So for example, we could add LegalZoom, Play MCP into a QuickBooks, Gusto, tons of different things that you could connect to inside of Claude.
19:45Now let's say there's an app that you wanna connect to that's not inside of this list that Claude and Anthropic give you. I'm gonna show you a hack right now that will basically allow you to connect to different applications that aren't inside of this. So I personally use something called Zapier MCP, but you could use really any MCP server.
20:02And what I'm gonna do is I'm just gonna click on start building here. If you already have an account, great. If not, make sure to sign up for one.
20:09And this is really only if there's apps that you wanna connect to that aren't inside of Claude. Otherwise, you don't really necessarily need to use this. First thing you're gonna do is you're gonna come over to the left hand side.
20:19Click on new MCP server. You're gonna click on Claude Cowork as your client. As you can see, I have a couple of different tools here in different apps that I've personally connected.
20:29And in order for you to connect some, you just come here and type in any of the apps you use. So, for example, if I wanna connect to school, I could just click on this, and I could select all of these different tools here. And from here, I could just click on add tool, and then we will see it populated right here.
20:42Once we add all of our apps we wanna add, we're gonna come over to connect and click add to Claude, and then we're just simply going to connect it here. There should be a button, and then we will have access to all these apps.
20:53Then you could come over to Zapier right here, and you could see all the different tools we have as well as the permission levels. If you don't wanna have to approve this before every single time that it calls this tool, you can just come here and select always approve. However, I would suggest being weary of doing that.
21:09Really only do that for specific actions that you know, you know, won't cause problems if it's off doing its own thing. I wanna quickly show you a connector I have that is pretty crazy. It's paired with some skills that I created inside of Claude.
21:22So let me show you right now. Basically, if I come over to my Instagram, I wanna show you these different Instagram carousels that I've been creating. Like, take a look at this.
21:30Five Claude code skills I can't live without. You probably thought, like, Brock, did you make this in Canva? How did you do this?
21:36And this was all generated inside of Claude with a skill I created, and this literally takes me, like, two minutes to build. I literally just have to give a prompt to it, and it will generate these automatically. Not only that, though, it will even come up with the caption and post these automatically for me inside of Claude.
21:54And in order to do that, I'm using something called Platato. As well as if you want access to this Instagram carousel skill that I'm talking about, I actually showcase it in my school community, so make sure to join. There's a link in the description as well.
22:06So, basically, how this tool works is you could come here and you could select any of these different social media platforms that you wanna connect. I personally just have my Instagram connected, but you can actually cross post all of these posts across Twitter, Instagram, LinkedIn, all these different platforms. So it's an easy way for you to repurpose content.
22:23Next, you could see I have a calendar here, and this is where I have my different Instagram carousels scheduled to go live. One cool thing that might actually be helpful for you is there's also templates in here. So let's say you wanna generate specific infographics and wanna post this on LinkedIn, for example, you could actually take this, use this as an example, plug it into Claude Cowork, and, you know, turn that into a skill so it will automatically generate it in that style or even, this tweet card.
22:48Tons of people make posts like this that go really viral. So you could literally just, like, repurpose YouTube video or something and then use this template to create content for you on autopilot. I'm just gonna show you real quickly how you could set this up inside of connectors.
22:59So what we're gonna do is come over to CoWork, and we could add custom connectors if there's a specific MCP server we wanna connect to. So I'm just gonna call this potato test. And then from here, I'm gonna come over to potato.
23:11I'm gonna come to API. So I'm gonna come over to where it says Claude desktop and Claude code.
23:17Copy this URL here. Then I'm gonna come back to Claude.
23:21I'm gonna paste that in, add it. And then since I already have this connected, it's not gonna work. But then it will just, like, pull you back over to Platato, and you just have to click connect.
23:29It will automatically connect it so you could begin scheduling posts on autopilot inside of Cowork. So that is connectors. There's tons of different use cases for this, so I suggest playing around with it and seeing what apps you could connect to that you use on a day to day basis.
23:43Moving on to concept number 16, and this is the Claude Chrome extension. And this is where you could have Claude go off and perform tasks inside of your browser, whether that is filling out forms, clicking on things, booking flights for you. This is one step below connectors.
23:59So if there is an app that you can't connect to, you could always use the browser capabilities, and I'm gonna show you exactly how to use this right now. What we will do is we'll come back to connectors, and then there's gonna be a Claude in Chrome button here.
24:12Make sure to enable this. And then next, what you're gonna do is you're gonna look up Claude in Chrome. So we're gonna come to this link right here.
24:18There will be a link in the description to access this. And what we're gonna do is we're just gonna add this Chrome extension. I already have it downloaded.
24:25Just make sure to download it. And then we could always access this right here. And from here, we could just give it a task and make sure to click on act without asking if you don't wanna have to continue to, you know, grant the permissions, and it will go off and perform different tasks for us.
24:40So in order to show you what I mean, I am inside of coworker, and I said, can you go on school and pull the last post in there? And the reason I'm doing this is because there's no way to connect school to my Claude in order to pull different posts. So what we can do is pull this up.
24:54And as you could see, this little red ring around my screen, I was not controlling my computer. It pulled this up, and it's literally going and finding that last school post for me. So it's able to actually browse the Internet and pull different things for me without me having to touch my computer.
25:10Moving on to concept number 17, and this is similar to the Chrome extension that I just showed you, but this is computer use. So this is a feature that they just released not too long ago, about two weeks ago, and Claude could actually use your computer not just inside of your browser.
25:26So if you have different files you wanna access or let's say you have a video that you need to upload to CapCut or to iMovie, it's able to go on your computer, click through the videos you have on your desktop, and then literally drag it and upload it into a application. To show you what I mean, let me give Cowork this task to find this video on my desktop and upload it to CapCut.
25:49I have the video file literally sitting on my desktop. So what it's gonna do is it's automatically gonna pull up computer use. As you could see, it's working on computer use here.
25:57I'm gonna click on allow for this session. And as you could see, again, there's this red ring around here, and I am not touching my computer. It's literally going and finding this video file, and it's automatically gonna add this to CapCut, which is my video editing software.
26:12Again, hands are off the keyboard. It just pulled up CapCut. I think it found the video, and it should upload it here quickly.
26:18And just like that, it actually added this video here. That is pretty crazy, the fact that it's able to go and basically access anything on my computer if we grant it access to. To.
26:27A key thing I wanna talk about here is connectors are the best way to go. It's the easiest way for it to perform different actions across our apps. Then browser use is probably the next best thing, and then computer use is probably the thing that I don't rely on much.
26:40But if I'm away from my computer, you know, I could have it go ahead and actually find video files for me. Because and the reason why I say this isn't the most efficient is because it's pretty slow. It's not really the best.
26:51It will probably get better over time. But as of right now, it's easier to just use connectors. Moving on to probably my favorite feature inside of Cowork, and this is scheduled tasks.
27:01This is basically what I've replaced n eight n in different automation platforms with. I use this to automatically trigger different actions throughout the day at specific times to go off and do different tasks for me, and this is built on top of the skills feature that I showed you earlier in the video.
27:17So for example, every single morning, instead of me having to manually give it a slash command of morning briefing, it will automatically go and invoke that skill and perform this task for me without me having to do anything. To show you exactly what I mean, what we could do is we could come over to co work.
27:34On the left hand side, you will see scheduled tasks. And right here are the different scheduled tasks I have running every single day. And, again, these are just basically automations that are running.
27:42So I have a daily wrap up skill, which basically is a way for Claude to tell me exactly what it did for the day, and then it will automatically generate this HTML dashboard for me. This is one of my favorite things, honestly, So that way I could see what did Cowork do for me today when I was away. Then I have other things such as school community, um, polls.
28:01So what this will do is it will automatically pull all the different posts and comments in my school community for me every single day. So then I could see what I need to respond to and kind of what's going on inside of there without me needing to be glued to my school community. In order for you to create a scheduled task, there's a couple of ways that we could do it.
28:19We could come over to scheduled and then add new task, give it a name, description, and then a prompt as well as choose a frequency. So we could do every hour, every day at 9AM, etcetera. We could basically just, like, customize this.
28:30Or what you could do is let's say you have a preexisting task that you're, you know, doing inside of co work. You could then just, like, say, that's great. Can you now run that every single day at 9AM?
28:40And it will automatically turn that into a scheduled task for you. And then if we go back to our projects, again, this is one of my favorite things is using projects because I could see all these scheduled tasks inside of projects. So this is anything YouTube video related.
28:53These are all of my scheduled tasks or automations that are running, and I could pull it up, see the instructions, run it, or change the instructions here. Next up for concept number 19, this is dispatch mode.
29:04So this is a new feature they released that allows you to control Claude on your computer so it could go and do all the things CoWork can do on your desktop, but you can control it now from your phone. So there's a little button here that's called dispatch.
29:17And, basically, what you could do is just type with it and say, hey. I want you to go grab the receipts from my receipts folder, and it will automatically pull it up on your computer and grab it for you. Then one thing to mention is you will see a dispatch section here inside of Cowork on your desktop, and you could see all the conversations, kind of a unified interface that basically shows the conversations that you could also see here on your phone as well as on your desktop.
29:41This is a feature that's really handy if I'm out on the go. Let's say I'm at the gym and I need to send my editor a video. I could say, hey.
29:49Go pull that YouTube video and upload it to Google Drive, and it will be able to do that. You could also trigger different skills that you have inside of dispatch. So this is a pretty powerful tool if you use it correctly.
29:59Moving on to concept number 20, and that is sub agents. This is basically a concept that allows Claude to do multiple different things for you at the same time with different agents or skills that you created in order to perform a complex task. So I like to think of these as .
30:18So let's say we give the request prep everything for my investor pitch. Agent one will then do the market research because it's trained on that. Say you have specific skills for market research, so we'll go do that.
30:29Meanwhile, agent number two is pulling your financials. Agent three is writing the narrative, and agent number four is building a slide deck. So this is just a really powerful way to combine all the different skills and agents you built inside of Cowork into one workflow in order to basically do more complex tasks for you.
30:47And there we have it, guys. That is the 20 Claude co work concepts that you need to understand as a beginner, and I broke them down very, very simply. If you guys want more content like this, make sure to subscribe to this YouTube channel.
30:59I cover all things AI for non techies to show exactly how you could leverage this powerful technology that could sometimes be very much of a headache to try to stay on top of. With that, if you guys want to join my school community, there is a link in the description below to get all the 50 plus skills I use every single day.
31:14I update it weekly, and there's some amazing entrepreneurs in there that are using Cloud Code Work daily, and there's some pretty cool use cases that you can use. With that being said, guys, thank you so much for staying the end of this video. I look forward to seeing you in the next one, and have a good day.
31:26Cheers.
The Hook
The bait, then the rug-pull.
Brock opens with the pain — you've watched the tutorials, you still don't know what an MCP or a sub-agent is. He promises that by the end of these 31 minutes, every Claude video you ever watch again will finally click. The bait is comprehensiveness: not one concept, all 20, simplest to most powerful, in order.
Named ideas worth stealing.
00:30list
20 Claude Cowork Concepts (simple to advanced)
- Workspace Folder
- Claude.md
- Global Instructions
- Memory
- Context Window
- Multimodal
- Web Search
- Extended Thinking
- Artifacts
- Projects
- Bash Tool
- Skills
- Slash Commands
- Plugins
- Connectors
- Chrome Extension
- Computer Use
- Scheduled Tasks
- Dispatch Mode
- Sub-Agents
Numbered glossary that walks beginners from filesystem basics to multi-agent orchestration.
Steal forany tool with a confusing ecosystem — make 'Every Supabase concept', 'Every Trigger.dev concept', 'Every $6 Stack concept'
05:08concept
Claude.md vs Global Instructions (project brain vs permanent identity)
Project-scoped instructions live in the workspace folder (Claude.md); identity that follows you everywhere lives in Global Instructions. Clear separation kills 90% of confusion.
Steal forexplaining any system with multiple scopes of config — env vars, settings.json layers, etc.
25:10model
Connectors > Browser Use > Computer Use (hierarchy of access)
- Connectors (native API)
- Browser/Chrome extension
- Computer Use (full desktop)
Always reach for the highest layer first — connectors are fastest and most reliable, computer use is slowest and most error-prone.
Steal forany tool-choice framework where reliability beats power
How they asked for the click.
VERBAL ASK
30:50newsletter
“If you want all 50 plus skills I use every single day, there is a link in the description below to join my school community.”
soft, value-stacked — referenced 4+ times throughout the video. No hard pitch block. Description holds the bit.ly to the paid course and DWY offer too.
MENTIONED ON CAMERA
FROM THE DESCRIPTION
PRIMARY CTAWhere the creator wants you to go next.
AFFILIATECommission earned if you click.
OTHER LINKSAlso linked in the description.
Visual structure at a glance.
