NEW Claude Parallel AI Agents Update is INSANE!

There's a brand new update from Claude with the new desktop app and this has totally changed how you can run AI agents directly with Claude. So I'm going to show you exactly how this works step by step. I'm going to run through it together. So what you can see here is that we've got multiple chats right inside this project.

And so with Claude now you can have multiple different tasks running at the same time, working across different projects and using agents in parallel. This is literally a whole new setup designed to have parallel AI agents. So, for example, here we can have multiple agents, one creating my own version of OpenClaude, which you can see in this chat. We've also got, for example, a new website that's been built out inside this section.

And then we could even, for example, like create our own AI SEO keyword tools inside another chat. And so the main point here is like Claude literally built something for parallel agents. So this is a new update that just dropped. You know, they've redesigned Claude's desktop essentially so you can run Claude sessions side by side from one window of a new sidebar to manage them all, which is pretty cool.

You can also preview stuff. You've got diffs, you've got terminal on the right hand side. And so what is this about and what even is this? So imagine you have like a really big project, right?

You could do it all by yourself, but that would take forever. Or you could call 10 of your friends, split the work up and finish it in 10 minutes. That's exactly what parallel agents are, right? Instead of one AI doing everything one step at a time, you send out multiple agents at the same time and they'll work together simultaneously.

And Claude, that's the AI made by a company called Anthropic, can run these agents for you. It's like having an army of robot assistants all working at once. So, for example, if we run a new session here, we can set up one here and we'll say, OK, create. We'll put brand colors here and we'll say change this to my brand colors.

And we'll change that website I just showed you a minute ago to my brand colors. And then we've got, you know, this has been here. We can say, OK, add more content plus blogs over here. Right.

You see how we've got all these agents working in parallel across the same project. Right. These two agents here. So who made this and why should you care?

Well, Anthropic is a company that built Claude. They're one of the most respected AI safety companies in the world. And Claude is a flagship AI and it's genuinely one of those powerful AI tools available right now. So that quote means the harder the job, the more powerful parallel agents become.

Right. This is a quote, by the way. Multi-agent frameworks increase Claude's ability to complete difficult long horizon tasks. So this isn't just a cool trick.

It's really a fundamental shift in the way that AI agents get work done. So you can see here, for example, whilst I'm talking to you, it's adding six more blogs over here. And over here, it's actually improving and changing the colors on the actual page. So why does this matter so much?

Why does speed matter so much? Well, every second you wait for an AI to finish a task is the second you're not growing your business. Right. Normal AI works like a single checkout lane at a supermarket.

Parallel agents work like opening 10 checkouts at the same time. It's more efficient. And according to Anthropic's research, tasks that used to take hours can be compressed into minutes. That's not an exaggeration.

That's the actual mathematical reality of parallelism. Now, how does it actually work? This is a simple version. So you've got, for example, the orchestrator, the boss.

Right. You've got the main core agent, which is working over here. And you also have the sub-agents, and these are working together. Right.

We can build a team of them at any time with these sub-agents. So let me show you an example of what this could do in reality. So we could have one agent that searches, you know, if we were doing research, one agent could search topic A, another B, another C, in different threads here, all at the same time. So you get a full research report in a completely fraction of the time.

Right. So we could run a new session here, and then we could open up a new folder. So we'll create a new folder here called, for example, Claude Agents. And then we can spin up a new task, right, inside this folder.

So we can say, OK, research DeepSeek. Right. We'll trust our workspace, I'll set up a new one. And then we'll do a new session inside the same folder.

And we'll say, OK, research AI agents from Kimmy. And you see how it's running the same task. Sorry, it's running in the same folder, multiple different tasks. And these are all working together, as you can see right here, in parallel.

That's what we're looking at here. You could do this for content creation. You could have one agent write the intro, one write the body, one write the conclusion, one check for grammar, and it's all done simultaneously. You could have, for example, coding projects.

You could have one agent write the front code, one write the back end, one write the test, all running in parallel. You could have two different projects working over here as well, like you can see. You can have one agent reading file one, one file two, one file three, all working together. Now also what we have is Claude Routines, right.

And so, basically routines are like kind of scheduled tasks where you can have these AI agents go off on a schedule, right. So, you know, what is a routine? Well, you know how your alarm clock goes off every morning without you touching it? Basically a Claude code routine works exactly like that.

But instead of you waking up, it does your work for you. So a routine is a safe set of instructions you give it to Claude, right. You set it once and then it runs automatically on a schedule when something happens or whenever you tell it to. And the best part here is like it keeps working even when your laptop is closed or your phone is off just through this routine.

So we can create a new one like this and then we can run a remote routine right here. And Anthropy actually says routines execute on Anthropy-managed cloud infrastructure so they keep working when your laptop is closed. So the cool thing about this is we can have routines running in the cloud even when I don't have my computer on, which is amazing. Why does this matter?

Well, most people are doing the same tasks over and over every single day. Checking emails, reviewing code, reading reports, organizing files. That is wasted human time. So routines hand all of that repetitive tasks to Claude permanently.

And you do it once, set it up, and Claude does it forever. That's what we're doing here. So what does a routine actually contain? Well, every routine is made up of three core ingredients.

You've got the actual prompt, which you can plug in over here and you describe what you want it to do. And it tells Claude exactly what to do every time the routine runs. The more specific your prompt, the better your results. It's kind of like a recipe, right.

Vague recipes make terrible food. It's the same with prompts. And then you've got repositories, right. So a repository or repo is basically a folder where your code or files live.

So you can select a repository here by installing GitHub. And Claude connects your GitHub repos and works directly inside them. And then you've got your connectors, right. So connectors like Slack or Google Drive or Linear, they connect to your Claude.

And then this lets Claude read from and write to those tools as part of the routine, which you can do down here, which is pretty crazy. Now there's three types of triggers. How routines start. So a trigger is what kicks a routine off.

And there are three kinds. You've got the schedule type, right. So this runs your routine at a set time, like an alarm clock. And you can trigger it through this section here.

You've got three different triggers, as you can see. So you can set this to run hourly, daily, or on weekdays or weekly. Example, like every Monday morning at 9 a.m., Claude reviews all the new pull requests. You set this time once, and then Claude shows up every single time without fail.

You've also got API calls, right. And this is for the more technical folks. But basically an API call means another piece of software can tell your routine to run. So an example could be monitoring your system, and it detects an error.

And it automatically pings Claude to investigate and fix it with no human needed. And the computers can just talk to each other, and Claude goes off to work. And then finally you have GitHub events, right. This runs your routine when something specific happens in your code.

So an example could be someone opens a new pull request, and Claude automatically reviews it and leaves comments. Like a GitHub trigger starts a new session automatically when a matching event occurs on a connected repository. That is the anthropic documentation. Now, let's talk about real world examples.

So, for example, every weeknight Claude could read all the new issues in your project tracker. It labels them, assigns owners, and posts a clean summary to Slack. Your team wakes up to a tidy, organized queue without anyone doing it manually. You can also fix bugs automatically.

So your monitoring tool could detect an error. It calls Claude's routine instantly. Claude finds a bug, connects it to recent code changes, and opens up a new draft to fix for your team to review. So you've got a team of on-call engineers who can review a solution instead of starting from scratch at 3am.

And your automatic code review. So every time a developer opens a pull request on GitHub, well, it can check for security issues, performance problems, and style mistakes. It can leave inline comments so human reviewers can focus on the bigger picture here. By the way, we've got that research automated here, and also automated right here.

Pretty crazy. And you've got post-deploy verifications. After every deployment, Claude can run smoke checks on the new build. It can scan for errors, regressions, and problems.

And it can post a go or no-go before the deployment window closes. And then it can keep documents up to date. So every week, Claude could scan all the code changes from the past seven days, find documentation that's now out of date, open, update, pull requests automatically, and then your docs and your documentation for any sort of project just never goes stale again. Now, let's talk about safety here.

So routines run completely on their own. No one is watching during these runs. That means that the prompt you write needs to be very clear and very complete. Claude can run code, push to GitHub, and send Slack messages and interact with real tools.

And mistakes can have real consequences. Anthropic's actual documentation warns about this. So they say routines can run autonomously as full Claude code cloud sessions. There is no permission mode picker and no approval prompts during the run.

So always limit what the routine can access to only what it actually needs. And by default, Claude can only push the branches to start with Claude forward slash. And that just protects your main code base from accidental changes. So who can use this?

Well, routines available on Pro Max Team and Enterprise plans. You need Claude code on the web enabled. You can create and manage them at Claude routines. And each routine belongs to your personal account, right?

It's not shared with teammates. Everything Claude does shows up as you. Commits, pull requests, Slack messages, et cetera. So that's basically it for the new update.

If you want to get all of the training on this, all my guides, et cetera, you can get that inside the AR Profitable Boardroom. Link in the comments description. Let's go to the arprofitboardroom.com. You can see that we have a full video tutorial and a step-by-step guide breaking through exactly how to use this, along with my new trainings on OpenClaude and Hermes and all this other new cool stuff that's come out recently.

This is inside my AR Profitable Boardroom community where you can ask questions, get help and support, learn how to save time and grow your business with AI. You can post inside the community and get questions and help whenever you want to. You can also join the weekly coaching calls where we get deep on sort of Claude Code and how to use Claude and how to scale with AI agents. You can get all of my best trainings inside the classroom.

And then also you can meet people in your city who are using Claude and Claude Code just like you are, right? Which means you can meet up with them, you can connect with people, you can DM them if you want to. And that's all inside the AR Profitable Boardroom. Link in the comments description or just go to the arprofitboardroom.com.

Thanks for watching.