# Claude Code just shipped the monitor tool ## Метаданные - **Канал:** Ray Amjad - **YouTube:** https://www.youtube.com/watch?v=MpSf7EN5dhc - **Дата:** 10.04.2026 - **Длительность:** 7:53 - **Просмотры:** 17,887 ## Описание Level up with my Claude Code Masterclass 👉 https://www.masterclaudecode.com/?utm... Use LASTCALL for 35% off lifetime (before lifetime plan is retired on Friday 24 April 23:59 PST) My Claude Code newsletter 👉 https://www.masterclaudecode.com/news... My Claude Code coaching 👉 https://www.masterclaudecode.com/cons... I've never accepted a sponsor; my videos are made possible by my own products... —— MY CLASSES —— 🚀 Claude Code Masterclass: https://www.masterclaudecode.com/?utm... —— MY APPS —— 🎙️ HyperWhisper, write 5x faster with your voice on MacOS & Windows: https://www.hyperwhisper.com/?utm_sou... Use coupon code YTSAVE for 20% off 💬 AgentStack, AI agents for customer support and sales: https://www.agentstack.build/?utm_sou... 📲 Tensor AI: Never Miss the AI News on iOS: https://apps.apple.com/us/app/ai-news... on Android: https://play.google.com/store/apps/de... 100% FREE 📹 VidTempla, Manage YouTube Descriptions at Scale: http://vidtempla.com/?utm_source=yout... ————— CONNECT WITH ME 🐦 X: https://x.com/@theramjad 👥 LinkedIn:   / rayamjad   📸 Instagram:   / theramjad   🌍 My website/blog: https://www.rayamjad.com/ ————— ## Содержание ### [0:00](https://www.youtube.com/watch?v=MpSf7EN5dhc) Segment 1 (00:00 - 05:00) Okay, so we have a brand new tool inside of Claude Code called the monitor tool. And I'll be going over exactly how this works and how you can be using it in your own workflows. Now, essentially what the monitor tool allows for is a real-time event streaming in Claude Code from background processes, whereby events that you care about are being delivered from a background process to the main session of Claude Code, which means that tokens are only used when events you care about actually happen, rather than continuously. So for example, you can be working on a new feature and say to Claude Code, start the development server and watch for errors whilst I work on the auth feature. It will then set up a monitor in the background and the dev server will be running. And whenever an error happens, like you may have triggered an error by clicking around on the local preview version, then that would go directly into Claude Code and then it would like identify what that issue is. So if I gave that command to Claude Code right now and pressed enter, then you can see that it's set up a monitor right over here. And the first monitor event came through. So we have no errors right now. And if I then go down, click on this, I can then see that we have the monitor running right over here. And these are the monitor details. And this is what Claude Code is looking out for. So it's looking out for error, warn, failed, ready, and a few other things. And then for the shell, this is the npm run dev that is happening. So I just opened up the local server. And you can see that it gave us a warning which appeared in the monitor right over here because it triggered warning and then went back into main session. Now, for example, if I was on the website and then I triggered some kind of error or something, then that would immediately go back to Claude Code into this Claude Code session as a brand new monitor event. So I could then start working on a feature. And then when I test in the browser, if an error does trigger behind the scenes, then Claude Code will be immediately notified with the error. So I wouldn't have to tell it every single time, hey, there's an error that just happened. Now let's do a quick comparison between the other ways of running these commands. So previously you could have shell commands run in the foreground. So you could say run the tests and then it would run the tests in the foreground. But that would mean that you couldn't enter in any more commands and have them run immediately because they would be queued since you have something else running in the foreground. If you had this running in the background instead, then that would run in the background all like 47 tests. And only once a background command exited would it actually notify the main session, like, hey, these are the tests that passed and failed. So in this case, we would only get one notification at the end and not know about any failures until all 47 tests had finished. But in the case of a monitor, if I said run the tests and like use a monitor, then it would start running the test-like command. And as tests are passing and failing, that would be passed back into the main session. So I could specify like run the tests and monitor only for failed tests. In which case only failed tests would pass back into main session. And then Claude Code could start immediately diagnosing those issues with the failed test. So if you had tests that were running for like 10 minutes, then Claude Code would be diagnosing them as each one fails instead of waiting until those 10 minutes had passed and then diagnosing it. And that is essentially because anything that matches this filter that was defined in the monitor is being passed back into the main session. Now the way that it works behind the scenes is that we have 4 parameters for the monitor. We have the description of what it does, the command itself, and each line that the monitor detects and filters for is considered one event. And then the timeout and whether it is persistent. So essentially you would describe what you want to watch. Claude Code would then write the command with the right filter for it. That command would only print when something happens and that thing happening would be considered one event. And each print of that would be considered notification that Claude Code can then react to. So you can kind of consider this to be a security camera that only alerts on motion. And the reason why this is more efficient is because previously you may have had the shell running in the background and then Claude code would have to like poll every 30, 60 seconds to see if there are any errors. And every single time it checks, is it ready? That would be consuming a little bit of tokens that can add up a lot over the long term. Whereas with the monitor, only when things match the event does it consume tokens. Okay, now you may be wondering like, how is this different from /loop? Now /loop is time-driven, which means that it fires a prompt every N minutes. So like firing, hey, check the development server every 2 minutes. And each iteration is a full API call. But this is a much more token-efficient strategy whereby this is event-driven instead. So the script that Claude Code writes for the monitor, it would only filter for those events that are relevant, which means that there are basically zero costs when no event is being triggered. And when you do have a monitor, then you may essentially have two different types of commands running. You may have a stream filter or a poll & diff filter. So stream filter would be kind of like this one that I have running over here with the npm run dev. That is essentially log tailing. Essentially here, events are being triggered in real time. Whereas polling may be like you're checking an API endpoint at a certain interval, and then that monitor would be only delivering ### [5:00](https://www.youtube.com/watch?v=MpSf7EN5dhc&t=300s) Segment 2 (05:00 - 07:00) events into Claude Code once a certain threshold is met. So this can be more handy for situations like remote APIs and stuff like that. So let's actually go through a couple examples. Firstly, you may have a file drop watcher. So you may want to watch a shared folder for new files and process them as soon as they arrive. So that monitor would be watching that. And as soon as that event is triggered, because that file was added, then Claude Code would be woken up with that event and then start acting on immediately. You could also do something like setting up a monitor for an API endpoint. So for example, you may be pulling a price API for a stock price, and then Claude Code would only be notified once that price drops below a certain point. And that would go back into your Claude Code session, and then Claude Code can do any actions based on that. And I know some people have been using Claude Code recently to like basically buy and sell stocks, which does seem like a pretty fun thing to do, and I may have to try myself. And this monitor tool can be pretty useful for that as well. And next up, you may have some kind of business workflow yourself that the monitor can be notifying Claude Code of. So you may have launched a brand new sale and you expect to get a lot of sales, over the next like hour, couple hours, you can basically get a monitor to do a polling in the background. And then that poll would only drive events into Claude Code whenever something unusual happened as defined in the code over here. So this can also be handy for like launches for brand new products or like you just deployed a change to production and you want to monitor it for the next like 2, 3 hours or something like that, just to make sure the error rates certainly don't spike. And as soon as they do, if you were away from your desk, then Claude Code can quickly diagnose that issue because you said something like, hey, can you monitor this for the next 3 hours, our logs, and then notify me when the error rates spike above like 10 errors per second or something like that. So I would recommend trying this yourself. The main things that I will personally be using it for is the Vercel deploy and also running test suites as well. But sometimes I also deploy servers for clients. And I think in situations whereby we made a pretty critical change and then something went to production, we may want to monitor it for the next couple of hours. But essentially the main benefit here is that rather than wasting a lot of tokens because Claude Code is polling, it becomes more event-driven where it only reacts to things that happen that you care about. Now, if you do like this kind of stuff and you want to learn to be a Claude Code power user like me, then you may find my Claude Code Masterclass to be helpful. It is linked down below and I will soon be removing the lifetime access, so it will only be yearly. So if you do want to get lifetime access, then there will be a small coupon code down below for you to get lifetime access at a small discount before the lifetime access is completely removed. Now, some people from some of the world's biggest companies have already taken this and have gone on to be the best Claude code users at their companies. So you may also find it helpful for that reason too. --- *Источник: https://ekstraktznaniy.ru/video/48855*