{
  "video_id": "CSgbgtmhkAk",
  "channel_slug": "krishnaik06",
  "channel_handle": "krishnaik06",
  "title": "2- How Does Claude Code Work- Installation And Quickstart With IDE",
  "duration_seconds": 1974.0,
  "url": "https://www.youtube.com/watch?v=CSgbgtmhkAk",
  "upload_date": "",
  "transcript": "Hello guys, so we are going to continue\nthe discussion with respect to the Cloud\necosystem. And we are going to start\nwith the first product that is nothing\nbut Cloud Code, okay?\nNow let me tell you about Cloud Code.\nThis is becoming a very popular product\nnowadays because many many companies\neven in the enterprise level they are\nbasically using the enterprise version\nof Cloud Code.\nSpecifically to develop the product to\nbuild the products from scratch, to\nmaintain the products and even for the\nshipment purpose, right?\nSo as you see in the future interviews\nalso now,\nyou know, in companies they are\ndefinitely going to ask you whether you\nknow how to work with the help of Cloud\nCode because definitely the productivity\nlevel is quite high.\nNow when we talk about Cloud Code, many\npeople will just think that are we just\ndoing wipe coding with the help of Cloud\nCode?\nIt is not that guys. There are a lot\nmany things that we really need to learn\nfrom Cloud Code, okay?\nSo again this will be a series of\nvideos. This is just the first part of\nCloud Code and in this I have designed\nan agenda. So first of all,\nsince we are going to start completely\nfrom basics\nfirst of all we'll try to understand how\nCloud Code works, okay?\nThen second we are going to see the\ninstallation process and third we are\ngoing to go ahead\nquick start how you can integrate with\nyour ID, your favorite ID whether it is\na Visual Studio Code or Cursor or any\nother IDs as such.\nAnd then how you can go ahead and start\nthe session, okay? So everything I will\nbe discussing about step by step we'll\ngo ahead, okay? So the first thing what\nwe need to understand is that how Cloud\nCloud Code actually works, okay? And\nthis is something really important for\nyou all to understand.\nUh\nSo here I've just drawn some diagrams.\nI'll make you understand what exactly it\nis. But whenever we talk about Cloud\nCode, right? What exactly it is? It is\nnothing but it is an agentic coding\ntool.\nOkay? Agentic coding tool.\nNow this term is super important because\nwhenever we talk about agentic coding\ntool, right? This will be interacting\nwith your project. That basically means\nit will be able to read your code base,\nit will be able to edit the files, it\nwill be able to run the command, it will\nbe able to integrate with your\ndevelopment tools, whatever development\ntools you are specifically using, okay?\nSo you need to understand that, okay,\nfine. We are trying to\nuse this as an form of agentic coding\ntool. Whichever projects you have given\naccess to, you know, you will be able to\ngive some instructions and Cloud Code\nwith the help of your approvals will be\ndoing some kind of task. It will be\nediting some of the files, it will be\nrunning some of the commands, you know,\nit will be integrating with your\ndevelopment tool and many more things,\nright? And the best part about Cloud\nCode is that this is available\neverywhere, right? When I say everywhere\nthat basically means it is available in\nyour ID.\nIt is available in your terminals,\nright? You can directly use it directly\nfrom the terminal, you can also use it\nfrom the desktop app, right? Now when I\nsay desktop app that basically means\nCloud Desktop, right?\nAnd even you can actually directly use\nit from browser.\nSo this all I will show you, right? As\nwe go ahead in the sessions\nwe will try to understand how we can\nactually use in\nObviously I've shown you with respect to\ndesktop, you know, I'll also give you a\nquick idea about desktop because once\nyou have downloaded, right? In the\nprevious session I've already told you\nto download this. So here it is. This is\nthe Cloud Desktop, you know? Directly\nyou can select a folder and you can give\nthe instruction automatically the coding\nwill also happen over here, right? So\nyou can use this also. But as a\ndeveloper what I would suggest you try\nto use ID and terminals. That way you\nwill be able to see many more things in\nterms of how the coding is basically\nhappening, you know, you'll be able to\nquickly review things, you'll be able to\nquickly update the unit test cases and\nmany more things itself, right?\nSo\n[clears throat]\nNow let's go ahead and understand how\ndoes Cloud Code actually works, okay? As\nI said,\nCloud Code is an agentic coding tool,\nokay? So it actually uses an agentic\nloop, okay? So here let me go ahead and\nwrite it down.\nIt actually uses a agentic loop. Now\nwhat does this agentic loop basically\nmean?\nLet's say you go ahead and give a prompt\nover here. Let's say the prompt is,\n\"Hey, go ahead and create a mobile app\nor a UI design for doing this, for\ntracking your calories, whatever\ninformation you have.\" That we specify\nin the form of a prompt, okay? Then what\nthis agentic loop module will do, it has\nthree main parts, okay? The first part\nis something called as gather context.\nOkay, so first of all it will try to\ngather the context from the project base\nitself, whatever context it has, okay?\nOnce it gather the context, whatever\ninformation it has, it will pass to the\nnext step and here it will go ahead and\ntake some kind of action. Now when I say\naction, it is basically going to do the\ntask based on the prompt that is\nbasically given, okay?\nNow once the task is basically\ncompleted, then we go to the next step\nwherein we verify the results.\nOkay? Now this verification results\nagain here\nhuman can also do it, even\nour Cloud Code will also be doing it,\nokay? So all these three right? And\nthese three will be continuously\nhappening in a loop unless and until we\nget the satisfied result.\nOkay, so let me again go ahead and\ncreate this loop.\nSo this will continuously happening in a\nloop, okay? So after\nthe verifying result is basically done,\nright? If the results are not that good,\nthen again it is going to go ahead and\ngather the context again, it is going to\ntake the action and again it is going to\ncreate the output, right? And once this\ngets satisfied, it will go to the done\nstage. That basically means the task is\nbasically completed.\nAs a human being since we are giving\ninstruction to the Cloud Code to\nactually work, you can interrupt over\nhere, right? You can interrupt\nover here as a human approval, you can\nadd context over here.\nAnd you can do multiple task. You can\neven change some of the instruction, you\ncan modify the prompt if you want, you\ncan add more information related to the\nprompt and you can basically do that,\nokay? So this is how the Cloud Code\nactually works. Now see, this is\nimportant, right? Let's say this is your\nproject.\nWhenever you start Cloud Code in your ID\nor anywhere, like let's say in your\nterminal and all, right?\nCloud Code.\nNow this Cloud Code will be having the\naccess of this entire project, right?\nNow having the access of this entire\nproject is basically having the access\nof some of the context related to this\nparticular project. And there's some\nmore information that you can save it in\nthe form of context. There we have\nsomething called as cloud.md file,\nright? This I will talk about. You can\nalso go ahead and create this. This will\nbasically have some information about\nthe project. And you know, this will be\nvery helpful. Even understand, let's say\nthat you are a new team member who is\njoining a team.\nLet's say you are joining a new team,\nokay? Now you really want to have a KT.\nSo what the team will do, they'll give\nthe entire project repo.\nNow just imagine you have the Cloud\ninstalled in that particular system,\nlike let's say in your system.\nNow what you can do is that you can just\nsay to the Cloud Code, \"Hey, please\nexplain me about the project.\"\nOkay, explain me about the project and\nit will be able to give you the entire\ninformation about the project itself,\nwhat it is.\nYou know, probably tomorrow let's say\ntomorrow they give you a new use case to\nwork on, you can quickly ask Cloud to\nprobably modify some of the use cases\nand tell or provide more information how\nit can be implemented, right? So this is\nhow it works, right? And this agentic\nloop plays amazing role with respect to\nthis.\nAnd the loop actually\nadapts to what you ask. A question about\nyour code base might be might only need\ncontext gathering, okay? A bug fix cycle\nthrough all the various stages, you\nknow,\nrepeatedly. Like it can it can probably\nrun this agentic loop unless and until\nthe bug is not fixed. So let's say when\nthis agentic loop is basically\nexecuting, there may be a bug that may\nhave got, right? It will also try to\nexecute, it will try to execute the\nscript in the sandbox and you get some\nkind of bug, right? And based on this\nagentic loop it will again recycle or it\nwill probably re-iterate or run this\nentire\nthree modules unless and until that bug\nis not fixed, right? So this is how it\nactually works, okay? And you are also\nthe part of this loop you should need to\nunderstand. You can interrupt at any\npoint, you can add context at any point.\nYou know, let's say that you are not\nsatisfied with\nany specific model, you can change the\nmodel also, okay? So everything is\nactually there. Not only model, you can\nalso change various tools that can be\nused right over here.\nIf you want some more additional context\nfrom web search, you can add that\nparticular tool also.\nSo in short, Cloud Code serves as an\nagentic harness around Cloud. It\nprovides a tool, context management,\nexecution environment that actually\nturns a large language model into a\ncapable coding agent. At the end of the\nday you actually get a very good coding\nagent.\nOkay? So this is how the Cloud Code\nactually works, okay?\nNow let's continue this particular\nprocess and let's see that how we can go\nahead and do the installation, okay? The\ninstallation process is very simple.\nOver here you can see that\nfirst of all, if you want to directly\ndownload it in the desktop or VS Code or\nanywhere, you can directly click on\nthis, okay?\nBut the first step of the\nyou know, Cloud Code is basically\ninstalling it in your system, right? And\nwhen I talk about pricing,\nsee there are three different plans,\nPro, Max. As I already told you in the\nprevious session that you can take a $20\nplan every month and you can start using\nthat. Through that you'll be able to\naccess the Claude code and all. Okay?\nSo, first thing first, what you really\nneed to do is that I will just go ahead\nand open the documentation.\nOkay. So, let me click the documentation\nover here. Now, with respect to the\ndocumentation, you have all the options.\nLet's say if you are a Mac\nMac OS, you have Mac OS or Linux or WSL,\nyou can install from this particular\ncommand. If you have Windows PowerShell,\nyou can use this particular command. So,\nwhat I will do since I have a Windows, I\ncan probably go ahead and do this. If\nyou have command prompt also, like let's\nsay you have Windows CMD, you can use\nthis particular command, like like a\ncurl. So, what you need to do is that\njust go ahead and open CMD and paste\nthis particular command, right? So, once\nyou do this, the installation will start\ntaking place. Okay? Since I have already\ndone the installation,\nthen it looks absolutely fine that it is\nyou know, I don't need to again\nreinstall it, but I've already done the\ninstallation. But, what you can do is\nthat you can directly go ahead and\nsearch for the documentation and over\nhere, based on your laptop or desktop or\nwhatever things you are actually using,\nyou can use that specific command. So,\nthe installation is very very simple,\nokay? And then you can go ahead and do\nthe installation. So, I have already\ndone it, so I'm not actually going to do\nit. Okay?\nNow,\nwhat I'm actually going to do, I'll just\nclose my uh\nCursor ID. I'm actually using the Cursor\nID. You can use VS Code also, but I'll\njust show you. Let's say that you're\nworking in some of the projects\nbeforehand, right? So, what you'll do, I\nwill provide my Cursor with the access\nof that particular project, okay? So, I\nwill try to open that project with the\nhelp of Cursor ID. You can also use VS\nCode. It is up to you, whatever uh ID\nyou want to specifically use, okay? So,\nI will just go ahead and this is my\nCursor, okay? Quickly, it is basically\nopening it over here. And let's say I go\nahead over here. I will open the folder\nwhich I really want to work. Okay?\nUh so, let's say I want to work with my\nproject, some of the projects that I've\nbeen working. There are many many\nprojects which I'm actually creating,\nuh which is actually just for my sake,\nyou know, I keep on seeing multiple\nthings and all. So, let's say that I've\nselected this particular folder, okay?\nSo, this is what I'm actually going to\nuse. Let's say I'll close this, okay?\nAnd then I will also close the terminal.\nOkay? So, let me close it.\nNow, I have opened the terminal over\nhere.\nJust to check whether everything is\nworking absolutely fine or not, what you\ncan actually do, I'll just clear the\nscreen, okay? I will just go ahead and\nprobably see see after the installation,\nI will go ahead and see whether my\nClaude command is basically using. You\ncan also directly open it from the\ncommand prompt also, right?\nOver here only you can just write\nClaude, okay? And you can check. But, my\nworking as I said that I want to give\nthe Claude access to my project. So,\nlet's say this is one of the project\nthat I had actually done related to\nLangChain. And this is just like my\nupdated LangChain with respect to the\nrecent version, all the notebook files\nthat I've actually recently created,\nokay? So, I will just go ahead and write\nif the installation is done completely,\nI'll just go ahead and type Claude. Once\nI type Claude, you can see that it is\ngiving me all the specific information.\nOkay?\nSo, it is saying me that choose the text\nstyle that looks best on your terminal.\nI'll I'll just go ahead and choose dark\nmode because I like dark mode, okay? I\ndon't like light mode or color blind\nfriendly, this all things. I can adjust\nwith dark mode. So, I'll just go ahead\nand select this.\nNow, it is telling me select login\nmethod. Now, you need to have either of\nthis Pro Max, Team or Enterprise. I\nsince I use this Claude code in my\ncompany, so I have the Max version. So,\nwhat I will do, I will just go ahead and\nclick on first\nand it will go ahead and authorize or do\nthe authentication from the website.\nOkay? So, I've authorized I've given the\nauthorization, okay? Build something\ngreat. You're all set for Claude code.\nYou can now close this window.\nNow, I will just go ahead and close\nthis. Now, you will be able to see that\nlogged in as krish naik06@gmail.com.\nPlease enter to continue. So, here I\nwill click on and I'll press on enter,\nokay?\nNow, here you can see that there some\nsecurity notes which says that Claude\ncan make mistakes. You should always\nreview Claude responses, especially when\nrunning code. Due to prompt injection\nrisk, only use it with code you trust,\nokay? This is also there, so I'll go\nahead and create this. Then for the\noptimal coding experience, enable the\nrecommended settings for your terminal.\nShift plus enter for new lines. Yes, use\nrecommended settings. I'll just press\nenter, okay? And this is how it has\nbasically started, okay?\nNow,\nyou know that since I have opened my\nproject over here and you can see that\nit is basically having the access of the\nspecific project. Okay? Now, here is\nwhat you provide prompt. So, the first\nstep that I told you, right? Over here.\nWhen we talk about this is the prompt,\nright? And this prompt you can basically\nprovide in the specific terminal now.\nOkay? And here you can see tips for\ngetting started. Run {slash} init to\ncreate a Claude.md file with\ninstructions, something something is\nthere. {slash} model picker now list\nmodels and all. So, let's see some of\nthe commands that is given. So, first of\nall, I'll see, okay, which models I'm\nactually using. So, I'll go ahead and\nsee this.\nNow, here you can see that there are\nmultiple models that is provided by\nClaude. One is Sonnet, then you have\nSonnet with 1 million context. Then you\nhave Haiku, right? The default\nrecommended is Opus 4.7 with 1 million\ncontext. This is most capable for\ncomplex work. When we talk about Sonnet,\nit is 4.6, best for everyday task. Let's\nsay you are doing some kind of task um\nwhich is\nuh like a normal generic task, you can\nbasically go ahead and use Sonnet. Then\nyou also have Sonnet with 1 million\ntoken context, right? So, here you can\ncontext basically means that much uh\nconversation history it can basically\nstore. So, here you have Sonnet 4.6 with\n1 million context, builds as extra\nusage, so and so. And there is also\nsomething called as Haiku. Haiku uh 4.5\nis fastest for quick answers. For\nsimple things and all, you can basically\nuse this, okay? Use {slash} fast to turn\non fast mode. So, by default, I will be\nusing Opus 4.7 with 1 million context.\nIt is more capable for any complex work,\nokay? So, I'll press enter, okay? This\nis basically my model. Now, understand\none very thing one important thing,\nokay?\nIn the first stage, you know, whenever\nyou see some kind of\num\nsome kind of projects, you know, it is\nvery much important that you go ahead\nand create this Claude.md file, okay?\nNow, what is this Claude.md file, I'll\ntell you, okay? So, let's say\n[clears throat]\nI will just go ahead and ask a question\nto the Claude code. I'll say that, \"Hey,\nuh please tell me\nwhat this\nproject is all about, okay?\" So, this is\nmy question. Now, when I'm sending this\nquestion, what Claude code is going to\ndo it? It is going to analyze this\nentire project repository. It is going\nto check each and every file to find out\nwhat this project is all about and it is\ngoing to give me a summary, okay? is all\nabout and provide me a summary. Okay,\nI'll just ask a question.\nOkay? Now, once I'm asking this specific\nquestion, what as I said, Claude code is\ngoing to analyze all this particular\nproject. See, it is basically listing\nall the directories, readme file, md\nfile, everything. And it is going to\ngain that specific context, okay?\nNow,\nsee over here you can see bash command.\nIt is also asking me to execute some of\nthe bash command to see some of the\ninformation over here because there are\nfiles inside this particular folder. So,\nI'll just go ahead and press one, okay?\nSo, here you can see it is going to go\nahead and test, check every file,\neverything. If there is an error, it is\ngoing to reiterate. So, it is basically\nrunning an agentic loop over here now.\nThis entire agentic loop is basically\nrunning based on the prompt that I've\nactually given, okay? So, here you can\nsee, right? So, here I have got the\nentire information. See, a hands-on\ntutorial report demonstration, the new\nLangChain 1.1 or APIs notably create\nagent and the middle system that\nreplaces older agent patterns. These are\nall the files. You can see intro to\nthis, this, model provide integration\nand all, okay? Now, I'll say, \"Can you\ncreate me\na uh\nClaude.md\nfile?\"\nClaude.md file. Now, see why I'm\ncreating this Claude.md file. See,\nin the previous question when I said,\n\"Hey, tell me about this particular\nproject,\" then Claude what it did, it\nbasically went through all the specific\nfiles. So, it had no context about the\nproject, right? No information about the\nproject. And it started scanning each\nand every files and it probably, you\nknow, generated all the information and\nit took some amount of time. Now,\ninstead what you can do is that if you\nsee on the top, right? Here what it\nsays, \"Run {slash} init to create a\nClaude.md file with instructions,\nsomething something is given, okay?\"\nThis Claude.md file plays a very\nimportant role, okay? It will have some\ninformation about the project, you know,\nso that it can provide a context to the\nClaude code.\nRight? Initially to provide some kind of\ncontext, okay? You can actually use the\nClaude.md file, okay? In the Claude.md\nfile, we'll have some information. So,\nlet me just go ahead and ask a question\nover here. \"What is the use of\nClaude.md file?\" I'll just ask question\nover here. Okay, this is also very good\nway of learning things. You can directly\ngo ahead and ask questions directly to\nthe Claude code, okay? And it will\nprobably trying to uh know, give you\nsome of the information over here.\nOkay? \"What is the Claude.md file?\"\nOkay, sorry. This I think it did not\nuh\nOkay, there is no Claude.md file. It is\nbasically checking uh in my folder even\nthough it is not. But, here it has given\nsome information. Claude.md is a special\ninstruction file that Claude code\nautomatically loads into context at the\nstart of every conversation in a\nproject. It's where you put persistent\nguidance you want Claude to follow\nwithout having to repeat it session,\nright? So this is really important.\nOkay, now what the code base is what it\nis for right typical content all this\ninformation. So I will take all these\nthings and I'll say and and remember\nthis Claude dot MD can be created in\nlocations also. First, you can also load\nthis in a project level\nover here within this project. You can\nyou can also create this particular file\nwith the name dot Claude dot local dot\nMD, right? And this is only for your\npersonal purpose, right? You want to\ngive more context to the Claude code,\nyou know? So I will go ahead and create\nClaude dot local dot MD file, you know,\nwith more information. I don't want to\nshare it that with the team. If you\ncreate Claude dot MD in the project\nlevel, this you can actually share with\nthe entire team, you know? So that team\nwill also be, you know, when they are\nusing Claude code, some basic context\ninformation will be available. And then\nyou can also save this in a global dot\nClaude dot Claude dot MD file, which\napplies to every project that you create\nin this particular system. So this\nlocation, where exactly it is, where you\ninstall it, okay? So I will just show\nyou where that particular file\nmaybe, you know? So I'll go to my C\ndrive.\nI'll say users. I'll say Win 10. And\nhere you can see dot Claude. And here\nyou can go ahead and create a Claude dot\nMD file. So this is where the\ninstallation is basically placed. So\nthis basically applies globally to any\nprojects that you create every\nRight? So with respect to any ID that\nyou use, if you are creating that Claude\ndot MD file over here, it applies to all\nthe projects.\nYou can also create project level and\nfor your personal level also you can\ncreate it. And remember this personal\none, you will not even commit in the\nGitHub, right? This dot Claude dot MD,\nsince it is in a project level, you will\nbe able to commit it, okay? Now I'll\ntell you, okay,\nso for your boot camp repo, a Claude dot\nMD here could be useful. Noting this is\nLangChain. This this is, please create\nagent all example targeted. So want me\nto generate one using the slash init\nskill. So I will just go ahead and write\nslash init, okay?\nAnd then it is just going to initialize\nClaude dot MD file. And this will\nbasically happen in my local project\nlevel, right? So automatically now\nyou'll be able to see based on all the\ninformation in this specific project, it\nis going to go ahead and create the\nClaude dot MD. And why are we creating\nit? It is just to make sure that it\nautomatically loads in the context at\nthe start of every conversation in a\nproject. So if I go ahead and close this\nparticular entire prompt and again I\nopen it, right? Again I open my command\nprompt and start the Claude\nconversation, it should be able to load\nthat particular context, okay? So now\nyou'll be able to see I have enough\ncontext now writing the Claude dot MD\nfile. Automatically you'll be able to\nsee in some time the Claude dot MD file\nwill be available over here. That's\namazing, right? So here I'm not teaching\nyou how to do white coding over here.\nI'm teaching you how\nproductively like in a very good\nproductive way how you can basically\nmanage a project, get more and more\ninformation out of it. So now it says do\nyou want to make this edit to Claude dot\nMD file? I'll say yes, okay?\nAnd automatically Claude dot MD file is\ncreated and here you can see that. Okay.\nAnd now I will reveal in the file\nexplorer.\nSorry. I will preview this. So I'll open\npreview and you can see how beautifully\nthis provide this file provide guidance\nto Claude code when working with Claude\ncode in this repository teaching\nmaterial so and so. All these commands\nare there you will see in this this\nthis.\nNow [snorts] just imagine whenever I\nstart any conversation, the first thing\nis that this Claude dot MD file will be\nloaded in the context. And for me, it\nwill be very very easy for me to run,\nright? Now this is one of the thing of\ngetting started over here. Now you can\nask any questions, you can create any\nfiles, okay? I'll say I'll go ahead and\nask hey what does\nwhat does\nuh main dot py file do?\nOr what is the code in\nor explain the code in\nmain dot py, okay?\nNow if I ask this specific question, it\nis just going to directly read this main\ndot py. It is basically going to give me\na summary like what exactly we do,\nright? And quickly you'll be able to see\nthat how beautifully all the information\nwill be coming in front of you, right?\nSo now you have started the session. So\nhere you can see, okay, there are two\npieces worth knowing the main functions\nso and so. Nothing is there exactly.\nIt's just a simple main\ndefinition of main function. You can see\nhello from LangChain updated so and so.\nAll the information is over here. Right?\nSo this is how you can actually get\nstarted, okay? Now let me talk about\nmore things over here and some of the\npopular commands that you can directly\nuse. So first of all I'll just go ahead\nand write slash model as I told you.\nThese all things are basically there.\nYou can click on escape. Now one more\nimportant see these all conversation we\nhad, right? If I just go ahead and write\nslash context. Okay. You can visualize\nthe entire context how it is saving. So\nhere you can see this much context is\nfilled, okay? How many tokens has been\nfilled from 1 million? System prompt 98\ntokens 0.9% system 25.5 15.5k tokens.\nMemory 63 tokens. Skills 1.1 tokens.\nFree space is so and so. Auto impact\nbuffer tokens. This is there.\nNow you may be thinking Krish, what will\nhappen if this entire token gets filled,\nokay? And that also I will talking about\nas we said that there are a lot of\nvideos that are going to come in this\nspecific series.\nThere we'll try to understand more and\nmore. Don't worry about the skills, how\ndid I create the skills? Everything I\nwill create and show it to you and how\nthings are basically done in this, okay?\nNow let me show one more thing, okay?\nNow what I will do I will delete this\nchat, okay?\nI will again open it. I will open my\ncommand prompt, okay? Now you can see\nthat when we are quickly starting\nsomething over here, right? And right\nnow you know that we have done a lot of\nconversation over here, right? A lot of\nconversation.\nI want to reload all the conversation.\nThe previous context\nI want to probably reload it. You can\njust go ahead and write Claude\nslash R, okay? So once you basically go\nahead and write it, you will be able to\nsee that quickly,\nyou know, it will say you project\nsummary overview request, okay?\nLangChain updated project overview 1\nminute ago, 1 day ago, all the\ninformation. So I will just go ahead and\npress on this.\nAnd you can see that all my previous\ninformation is basically loaded. Now it\nis being able to load with respect to\nthat specific context also, okay?\nThat's the most amazing part over here,\nright? So we have learned about multiple\nthings. We have learned about slash\ninit. What is the purpose of the Claude\ndot MD file, you know? Which all\nlocation you can basically go ahead and\nsave. You can also go ahead and make\nthis as Claude dot local dot MD so that\nyou can add more instruction just for\nyour project basis, right? And this is\nhow you can basically work with it,\nokay? So this is just a idea that I'm\nactually going to give you, okay? And\nstill we are going to see a lot of\ncommands. There is something called as\nslash agents. You'll be able to see that\nyou can also go ahead and create sub\nagents. And what is the use of sub\nagents? Sub agents main use is that you\nwill be able to have different memories\nfor the sub agents, different context\nfor the sub agents and you can basically\ngo ahead and create those kind of agents\nfor different different task, okay? Now\nlet me go ahead and ask more information\nas over here, you know? I'll say\nwhat all tools\nare there in Claude code? See, at any\npoint of times just keep on seeing the\ndocumentation. You find any problems,\nyou find you you're not able to\nunderstand, ask that question over here.\nAnd it will be able to give you\nsuggestions, you know? Suggestions with\nrespect to everything that you go ahead\nand create, right? So let's say tools is\none keyword that is specifically using\nClaude code, right? And then we'll try\nto understand what are tools. You can\njust directly go ahead and ask the\nquestion over here. So anything that you\nspecifically see, right? You should be\nable to see it in a very easy way, okay?\nAnd automatically this Claude code will\nbe giving you the answer. So see here\nover here you can see what are tools and\nwhat all tools are there in Claude code.\nClaude code tools falls into few\ncategories. Here's a full inventory\navailable in this.\nYou'll be able to read any file, write\nany file, edit, notebook edit, glob,\ngrep, shell execution. You can actually\ndirectly work with bash, power cell, web\nfetch, right? You can actually fetch\nfrom a specific URL. You can do a web\nsearch. You can actually create sub\nagents and then again launch specialized\nsub agents, general purpose, explore\nplan,\ntask create, plan work modes, you know?\nWe'll try to also understand about the\nwork mode.\nThen here you have MCP server. Like what\nall MCP server you can actually go ahead\nand integrate with it and all, okay? And\none more amazing thing is that, you\nknow, there is also something called as\nuh\nuh There are three different modes,\nokay? So if I press shift tab, there's\nthere is something called as accept\nedit. There is also called as plan mode\non, okay? So let's say that I want to\nprobably go ahead and work on a new\ntopic over here, okay? So let's say I\nwill say hey\nI want to create\nsome\nsome new\nJupyter notebook\non vectorless rag\nvectorless rag using\num\nusing Python\nand page index\npage index library, okay?\nPage index\nlibrary.\nOkay? Now I am actually doing this in\nplan mode. You just have to press shift\ntab. So once I give this new\nrequirement, you know, first of all\nthe Claude code is going to make a\ndetailed plan out of it, okay? So, I'll\nsay, \"Please make a detailed plan\nand save it\nin\nplan.md file\nso that\nwe can see it, okay?\nNow, I'll just go ahead and press enter.\nNow, see see the best part is that\nI've told that, \"Okay, regard related to\nvector less rag, I want to go ahead and\ncreate a Jupiter notebook, okay? And\nusing Python and page index library is\nthere. Please make a detailed\nplan and save it in the plan.md file,\nokay?\" So, now here, what it is going to\ndo is that you can see that it'll note\non plan final action plan mode extreme\nrestricts me to writing the plan at so\nand so. After you approve, I can also\ndrop a copy at plan plan.md in the repo\nif you want. Yes.\nSo, I'll plan this out. Let me research\nthis this this all the information, you\nknow, and by default, you'll be able to\nsee in this MD file will be there and\nhere also will be able to get it, okay?\nSo, I will just go ahead and do that. It\nis See, over here it has done the web\nsearch everything, right? Web search,\npage index, vectify, vector less rag,\nweb search with this. Tool usage is\nbasically happening. Web search is\nbasically happening over here. So, I'll\npress enter enter enter. It'll do\nmultiple web searches and it'll probably\ngo ahead and create a detailed plan.\nRight? So, this is how you'll be able to\nsee that it can call multiple tools also\nand that's the power of cloud code over\nhere, right? See, it is being able to\nfetch information for vectify, page\nindex, all the information. So, this is\nthe GitHub it is basically exploring\nnow. I'll just press enter. Okay, enter\nenter. Automatically, you'll be able to\nsee that you are able to go into get the\ninformation.\nThat's the beautiful part.\nUh it'll keep on asking.\nPlease make sure to give every\npermission you see it. That is where\nhuman in the loop approval is basically\nrequired. Okay?\nI'm just trying to show you like over\nhere how that MD file is basically\ngetting created and automatically this\nfile will get created automatically and\nyou'll be able to see that, okay?\nJust let's let's wait for some more\ntime.\nUh\npage index collab, page index, okay? So,\nit is doing all that see all these\nthings are basically happening with\nrespect to the engine tick loop.\nRight? And automatically, it should be\nable to give me the plan, okay? Bash\ncommand see it is going to go ahead and\ncreate a file over there inside my\nupdated LangChain. Okay? And uh you'll\nbe able to also see things are happening\nover here.\nOkay? So, learn run cloud continue or\ncloud resume to con- resume a\nconversation.\nIt is basically exploring these things\nand all, okay?\nSo, uh once this file gets created, I\nwill be able to show you this plan.md\nfile over here basically getting created\nand in the next video uh in the next\nseries of videos, we are going to\nunderstand the further things.\nUh right now, let the code keep on\nhappening and in the next video, we will\ncontinue with this, okay? So, I hope you\nlike this particular video. Uh this was\nit from my side. I will see you all in\nthe next video. Have a great day. Thank\nyou one and all. Take care. Bye-bye.",
  "transcript_chars": 33531,
  "ingested_at": "2026-05-15T16:44:02.514584+00:00",
  "source": "channel",
  "yt_meta": {
    "view_count": 11543,
    "like_count": 267,
    "channel_id": "UCNU_lfiiWBdtULKOw6X0Dig",
    "categories": [
      "Education"
    ],
    "tags": [
      "yt:cc=on",
      "claude code tutorials",
      "claude code demo",
      "claude code skills",
      "claude code subagents"
    ]
  }
}