# How To Use Claude Computer Use Agent For Beginners - Claude Computer Use Tutorial ## Метаданные - **Канал:** TheAIGRID - **YouTube:** https://www.youtube.com/watch?v=MUn6686Y17Y - **Дата:** 23.10.2024 - **Длительность:** 8:54 - **Просмотры:** 24,489 ## Описание Prepare for AGI with me - https://www.skool.com/postagiprepardness 🐤 Follow Me on Twitter https://twitter.com/TheAiGrid 🌐 Checkout My website - https://theaigrid.com/ 00:00:00 Anthropic Introduction 00:00:23 Docker Installation 00:00:49 Mac Options 00:01:10 Windows Downloads 00:01:42 Linux Download 00:02:03 API Setup 00:02:22 Key Creation 00:02:41 Key Storage 00:02:59 Security Warning 00:03:22 Docker Verification 00:03:55 API Configuration 00:04:35 Program Setup 00:05:18 Agent Introduction 00:06:06 Troubleshooting Tips 00:06:27 Error Handling 00:06:50 Prompt Guidelines 00:07:31 Screen Controls 00:08:10 YouTube Example 00:08:33 Final Demonstration Links From Todays Video: https://github.com/anthropics/anthropic-quickstarts/tree/main/computer-use-demo https://docs.anthropic.com/en/docs/build-with-claude/computer-use https://console.anthropic.com/ Check docker version - "docker --version" set ANTHROPIC_API_KEY=your_api_key_here Welcome to my channel where i bring you the latest breakthroughs in AI. From deep learning to robotics, i cover it all. My videos offer valuable insights and perspectives that will expand your knowledge and understanding of this rapidly evolving field. Be sure to subscribe and stay updated on my latest videos. Was there anything i missed? (For Business Enquiries) contact@theaigrid.com #LLM #Largelanguagemodel #chatgpt #AI #ArtificialIntelligence #MachineLearning #DeepLearning #NeuralNetworks #Robotics #DataScience ## Содержание ### [0:00](https://www.youtube.com/watch?v=MUn6686Y17Y) Anthropic Introduction anthropic computer use demo allows you to interact with Claude and AI model by simulating tasks on a desktop environment this demo is still in beta so it's essential to understand that some features may have limitations or be subject to change this tutorial is going to be for complete beginners that want to get started but have no experience dealing with any of the prerequisites that you need to get started firstly ### [0:23](https://www.youtube.com/watch?v=MUn6686Y17Y&t=23s) Docker Installation make sure you've installed Docker is an essential piece of software for this to run you can see right here all you'll need to do is to click download dock a desktop when you download Docker desktop there are actually five different options I'm going to explain what you need to do based on your current operating system the first one download for Mac Intel chip is the option for users who have a Mac computer with an Intel processor older Macs and ### [0:49](https://www.youtube.com/watch?v=MUn6686Y17Y&t=49s) Mac Options some models still use Intel chips and you should choose this option if your Mac is one of them download from Mac Apple silicone this option is for newer Mac that have apple silicone chips like the M1 or M2 chips these are Apple's own processors and you should select this option if your Mac uses one of these chips then we have Windows downloads for ### [1:10](https://www.youtube.com/watch?v=MUn6686Y17Y&t=70s) Windows Downloads Windows amd64 and this is for Windows computers with 64-bit processors this basically means that this architecture supports both AMD and Intel processors that run a 64bit version of Windows this is the one that most people on Windows will download and it's the one that I've downloaded now you've got download armm 64 which is essentially for computers that use armm 64 processors which are different from traditional Intel AMD processors and the armm processors are ### [1:42](https://www.youtube.com/watch?v=MUn6686Y17Y&t=102s) Linux Download commonly found in tablets some laptops and mobile devices if your Windows device uses an armm chip select this option then we have download for Linux and this option is for users running Linux as their operating system which is different from Mac OS or Windows if you're on Linux this is what you click next you'll need to head on over to this link this is ### [2:03](https://www.youtube.com/watch?v=MUn6686Y17Y&t=123s) API Setup console. anthropic docomo so first sign in on this link with your standard anthropic details that you use to sign in with Claude then you should be prompted with this screen ### [2:22](https://www.youtube.com/watch?v=MUn6686Y17Y&t=142s) Key Creation now once you're on this screen we can then get our API keys so you just want to click this button right here then what you want to do is you want to create your key so we go here and we click create key now you can see we can name our key for this I would advise you putting computer use because that is essentially what we are using it for ### [2:41](https://www.youtube.com/watch?v=MUn6686Y17Y&t=161s) Key Storage then for the workspace just click default and then click add once you've done this you'll then get a new API key now with your API key make sure that you keep a record of the key because you won't be able to view it again what you'll need to do is click the copy key button then we're going to open up notepad and then I'm going to ### [2:59](https://www.youtube.com/watch?v=MUn6686Y17Y&t=179s) Security Warning paste the API key in now you make sure that you save it in this document because there's nowhere else that you can view this key again now I'm sharing my key in this video just for tutorial purposes this key will be deleted by the time the video goes live but if you're someone that has an API key please make sure that you don't share it publicly or with anyone at all so now once we have ### [3:22](https://www.youtube.com/watch?v=MUn6686Y17Y&t=202s) Docker Verification our API key here let's just double check that docket is working before we start to run everything go over to your windows bar or whatever taskbar you have and then enter command prompt you can see this pops up right here and I'm just going to click this and now we can see this right here once you've got this on your screen we're just going to enter one small command that you just need to put in order to make sure that Docker is running correctly all you'll need to do is just paste what I've pasted in the description to check that your docket is working correctly once you paste this in automatically there should be this message underneath that shows you the docker version and the build if that ### [3:55](https://www.youtube.com/watch?v=MUn6686Y17Y&t=235s) API Configuration isn't the case put in whatever error message you get into chat G and diagnose until your Docker is working now that we have this we can now move on to the next step next you're going to want to set your API key so what I'm going to do now is I'm going to input my API key prompt the first part of that so what we will do is we'll put set anthropic API key then equals and then this is where we paste our API key so for example since I've just got my API key here I would paste this in and then I'd click enter and now that is my API key if you want to verify which API key you're using just to input this prompt and it will show you which API key they've currently set you can see ### [4:35](https://www.youtube.com/watch?v=MUn6686Y17Y&t=275s) Program Setup right now the same value was presented so I now know my API key was correctly entered in the first instance now that is all done the hard part is over all we'll need to do is enter the actual prompt to get this running this is where you'll input this so come to this link and just copy this part just click copy then come over to the system prompt and then put this in so for example here we'll just click enter and this will work and you'll see that the files should now start now I've done this and it will take a few minutes but as the files are starting you can simply wait once the files have been downloaded you're just going to need to click the link they've provided and then you'll get to the windows tab once you've ### [5:18](https://www.youtube.com/watch?v=MUn6686Y17Y&t=318s) Agent Introduction opened this up in the browser this is where you can now use the computer use agent now the agent is in an area that is completely contained meaning that it's a virtual workspace and not your own so now once you enter any kind of prompt into the prompt bar you'll see that it says running agent this means that anthropic is successfully running the agent and providing it with screenshots of the environment that is virtual so you'll have to understand that the environment is completely virtual you can see here in this recent example I asked it to go to Google and it simply sends screenshots and using those screenshots it positions the mouse performs a click and then moves to the next step this is the entirety of how the program works you can see that you're going to be also able to see exactly what's going on your screen ### [6:06](https://www.youtube.com/watch?v=MUn6686Y17Y&t=366s) Troubleshooting Tips with a left hand side view of the chat bar it's going to show you every single screenshot that occurs and where the steps are being placed now if you do face any kind of errors or any kind of rate limits what I would suggest you do is ensure that you're on the clawed paid plan often times if you're trying to use this on the free pan you might run out of trial credits very quickly now ### [6:27](https://www.youtube.com/watch?v=MUn6686Y17Y&t=387s) Error Handling sometimes this computer used demo can glitch and can perform instances where it gets trapped in an agentic Loop or might perform the incorrect click what I would do is I would stop running this and just rerun this and try again now it's also important to understand that this is early and in beta which means that sometimes things might not work as intended so it's important to understand ### [6:50](https://www.youtube.com/watch?v=MUn6686Y17Y&t=410s) Prompt Guidelines that you need to ensure not everything that you try will work one of the things that anthropics says to do in order to get the best performance out of the model is to ensure that you prompt clawed and then provide it with this prompt structure right here that says after each step take a screenshot and carefully evaluate if you have achieved the right outcome especially show your thinking I have evaluated step X if not correct try again only when you confirm a step was executed correctly should you move on to the next one now this is what they're saying to get the best quality outputs but it does work with simple prompts anyways and it also does say like some UI elements like drop downs and scroll bars might be tricky for exp to manipulate using movements so if you ### [7:31](https://www.youtube.com/watch?v=MUn6686Y17Y&t=451s) Screen Controls experience these issues where the model cannot navigate the web page try to tell the model to use keyboard shortcuts if you're finding that the agent has any trouble in this virtual environment if you click the top right area you can toggle stream control on you can connect to the virtual workspace and then you can close anything that might be making it hard for the agent of course that somewhat defeats the purpose but it's important to ensure that your AI agent having the tool use is working correctly remember this is a complete virtual environment and sometimes certain things on the web page can interfere with exactly what's going on and once you've done that you can then once again toggle the screen control and give control back ### [8:10](https://www.youtube.com/watch?v=MUn6686Y17Y&t=490s) YouTube Example to clae you can see that in this example what I've just done is asked it to go onto YouTube and find my most recent upload previously in many trials and errors the current agent was struggling because it couldn't get past this cookies page but when I went onto YouTube I asked it to find my recent upload you can see that it managed to take a couple of screenshots then manage to scroll down and find the latest ### [8:33](https://www.youtube.com/watch?v=MUn6686Y17Y&t=513s) Final Demonstration upload you can see that it gives me the information right here shows me how much views it's got and gives me a lot of information you can start to see that once you're able to do this with certain AI agents you're able to easily manage to automate many different systems and processes for you or your business hopefully this tutorial helped you get started and if there's any questions you have I'll respond to all of them in the comment section below --- *Источник: https://ekstraktznaniy.ru/video/13941*