All the AWESOME NEW Claude Code Features

Alright, so I'll be going over some of the Claude  Code features that came out in the last two weeks,   and I'll be going in reverse order from the  latest feature to oldest feature. There will   be timestamps in the description down below, so  you can skip ahead. Basically, getting started,

we have customizable status lines over here. So  if you run Claude Code in whatever folder you're   running it in, and you can write /statusline, you  can say something like "add the weather in London   with emojis", press enter over here, and then it  will fire off a built-in subagent in Claude Code   to add this to a status line. And that basically  means that just below this like input box over   here, it will show whatever information there is.   So you can see it proposes this edit over here. I   can press yes. And then if I exit Claude Code and  then run Claude Code again, and now you can see   over here it says "London Partly cloudy +17°C". I  can add more information to this, I can say "add   how long I've been in the session, the model  being used, the git branch, the folder name,   alongside the weather, the weather should be  last", press enter, and then it will make those   changes as well. And now if I press exit over here  to close Claude Code, and then I run it again,   then I should be able to see all these new pieces  of information. So you can see it says 0 minutes,   Opus 4. 1, main, vibe, and then Partly cloudy,  like this temperature over here. And you can   actually see what's happening behind the scenes  by going to your home directory over here,   pressing command-shift-dot on macOS, and you  can see the hidden folders, go to. claude. And   then you can see a. settings. json over here,  which calls the bash command for this like   bash script. And then you can open up the bash  script, press open with, Cursor, and you can see   all this information that Claude has over here.   And you can see where it's getting the weather   from, which is using this API endpoint over here.   And basically, you can get Claude to add other   things for you. So if you have an API endpoint,  you can get it to fetch your emails, for example,   like how many unread emails you have. You can get  it to fetch like how many dollars you've made so   far today on Stripe. And you can just like  import like other useful things that may be   helpful when you're running a coding session,  such as how many tokens you've used so far,   or like how many files that you've edited so far,  or how many lines that you've changed so far.

Anyways, the next thing that you can do is you  can run background commands. So you can basically   have Claude running and monitoring things in  the background. So for example, I can say,   "can you list the docker containers I have  running, and follow the logs of a random container   with the -f". And then it will ask if I want to  run this command. I can press enter over here.    And now you can see over here it says "1 bash  running". So this is running in the background,   and it's monitoring the output of this Docker  container. And I can also say "cd into nextjs   folder and run npm run dev" over here. And then  it will start running this as a like background   command too. So press yes over here. And now  you can see I have two bashes running. And if   I write /bashes over here, I can see the list  of bashes, which is the docker logs over here,   the cd over here. I can press enter to view  the like what's happening. Or I can press   'k' to then kill the like background command.   So if I press enter over here, it seems that   there's no output available. Or I can press  'k' to then kill the like command. And this   can be useful because you may have a development  server running locally, and you want Claude Code   to be able to like see any error logs that end  up happening by running its tool bash output.    So you don't have to always like copy over the  like error logs every single time, for example.

The next thing that you can do is you can use  at-mentions in slash command arguments over   here. So for example, if you make a slash command,  which is, for example, optimize, I just made this   one over here. And this is like a project-level  slash command. I can do at and then I can like   add a specific folder or file, for example. So  I can add this like OrganizationManager. tsx over   here. Press enter. And then it will only consider  this particular file when this slash command is   running. They also upgraded the Opus model from  4 to 4. 1. They also made it possible to specify

which subagent is running which model. And you can  watch my video on subagents that I previously made   using the link in the description down below.   So if you were to write /agents over here,   then run create new agent, for example. And I  can describe an agent to be an i18n translator,   keep the i18n files in sync, replace hardcoded  strings with their i18n equivalents. Press enter   over here. And then when making this agent, I  can choose whether it's running Sonnet, Opus,   or Haiku. And for this particular agent, Haiku  makes more sense. Like Opus is overkill and   so is Sonnet. So I can press Haiku over  here. And then if I open the agent file,   I can see it says "model: haiku" over here.   Something else they also did is adding hidden

files and folders to file search when you use  an at-mention. So if I write at over here,   then I can see. DS_Store, which is a hidden  file on macOS,. claude,. env, and so forth.

They also made it possible to at-mention specific  agents to invoke them. So I can do at and then run   @i18n. And I'll see this i18n agent that I just  made. And then I can run that particular agent.

They also changed one of the commands for  transcript mode. So if you don't know already,   if you press Ctrl+R when an agent is running,  then you can see everything that's actually   going to Claude. So you can see this bit of code  is like going over to Claude and also all these   other files as well. And this prompt over here.   If you press Escape now rather than interrupting   the agent, it just leaves transcript mode. Another  small improvement that I found useful is on macOS,

you can do Ctrl+V images. So if you run  Command+Shift+4 on macOS, you can take   a screenshot like this. And then you can press  Command+V into Claude Code over here. And then the   image easily goes there. So previously on macOS,  you would just have to drag and drop images into   the terminal when Claude Code is running. Now you  can just paste them. They also added support for

reading PDF files. So you can drag and drop a PDF  file into Claude Code and then just say "summarize   this", for example. And then it will like let you  know everything that's on the PDF file. And that   can be pretty useful for some people because  you may be given a feature spec in like a PDF   file or something from your product manager  and you can just put that into Claude Code.

And one more thing they added is allowing  you to specify which model is used with   which slash command. So for example, if  you go to your commands folder over here,   I have an optimize command. I can change this  to use Opus, for example. So at the very top,   if I write this over here, I can write "model"  and then I can just change this to "opus",   for example. So for some commands, you may want  to use more powerful models. Other commands are   kind of overkill and you don't want to eat up  all your context or use too many tokens. And you   can use simpler commands instead. And yeah,  that was basically every new thing that was   added to Claude Code in the last two to three  weeks. If you have found this video useful,   then do subscribe because I will be making  another one like this two to three weeks from now.