Introducing Augment Remote Agent: Parallel Autonomous AI Agents

--- type: transcript date: 2025-05-08 youtube_id: lBcsyGLh520 --- # Transcript: Introducing Augment Remote Agent: Parallel Autonomous AI Agents In this video, I'm going to be showing you Augment Code's new remote agent feature. If you haven't used Augment Code before, what it allows you to do is to have an AI coding agent directly within an IDE that you're already using. Regardless of whether you're using VS Code, Jet Brains, or Neovven, you'll be able to use this as an extension. In this video, I'm going to show you augment within VS Code. They also do have a free tier for this where you can try this out. And once you have it all installed, you should see within the extension marketplace within VS Code that we have this augment agent. to demonstrate the feature. What I'm going to do is I'm going to create a new repository here. I'm just going to create a simple Nex.js application. Once that's all installed, I'm going to go over to GitHub and I'm going to create a repository for this. I'm just going to call this remote agent. For now, I'll make this private. I'm going to create a repository. Once I have that, I'm just going to copy this code block here. And then once that's all synced, I'm just going to create my first commit message and push up that repo. And what I can do within the remote agent is since it has the context of our repo, we're also going to be able to have all of the different branches that we have associated with it as well. If that is the case, in this case, I'm just going to be working off main. You do also have the ability to customize your agents environment. The way that this is going to work is soon as we kick off a task to the remote agent is it's going to spin up a container. It's going to check out our GitHub repository. It's going to pull that down. It's going to index all of that content. And then once that's done, the agent is going to work through step by step all of the different pieces that are required for the request that we made. The first thing that I'm going to do here is let's update the homepage to just read developers digest. Let's remove everything else within there. It's going to do is it's going to allocate those resources for the container. It's going to spin up the container and then it's going to pull everything down. But what you can do is you can add in multiple agents in parallel which is really neat. What I'll say within this is I want to set up an API that returns hello world in JSON. I also want to leverage app router within Nex.js. Now a similar thing what it's going to do is going to take that query. It's going to spin up that container, grab all of those different resources and start to work through that task. Finally, what I'll say here is I want to set up a header as well as a footer on every page. Let's make new components and also make sure to update the layout so they're shown on all the pages. And so now the great thing with this is we have three parallel requests that you can almost think of this as giving different tasks to different employees. Within here we have a task that just finished. And if I go through everything and all of the different steps that the agent took. So it spun up the container, it cloned the repository, it made a new branch, it indexed everything that was up until that point. And then at that point we have the regular augment agent. Effectively the same agent that you'd have within your IDE. This is now just running on the cloud. What this is going to do, it has access to everything like the terminal as well as the ability to read and write files just like you might expect. We can see that it's going through and it's making all of those different changes. And what I can do within here is I can look at the remote changes. I can click through. I can see this diff here. It went ahead and it edited and we now have the update here. What I can do here is I can go and I can click to apply file by file or additionally I can click to apply locally for all of the different changes. So in this case there was just one change but now we'll see we have this page here where we have developers digest at the top of the screen. Now I see another agent has just finished and we have this API similar thing. It's going to spin up the container clone that make a new branch. It's going to index the repo and it's going to go and set off all of the different things that it needs to do within here. You can see it's leveraging the augment context engine. And the augment context engine is really powerful. You can have a repo with hundreds of thousands of files or tens of millions of lines of code and it does a really great job at finding the correct context that it needs to feed into the agent at any given time. There are also various tools available here. We can see it even did a little bit of research where it looked up the Nex.js app router documentation and then from there we see it's correctly setting up the API file within the app folder of API/hello. Once the agent is complete, it will give you a summary of the changes and in some cases it will even give you instructions on how you can potentially test it as well. Here I can open up those changes. I can see the diff of if it's edited files. In this case, it just made a net new file. And if I click to apply this locally, we should see it pop up within the app here. So we see this API folder and we also see this route. And so now if I go to our application and I go to the route of API/hello, we see this JSON object. Finally, we had the remote agent where we asked for a header as well as a footer on every page. Within here, we can see it read through the layout file, the page file, the global.css. It created a plan, and then what it did is it created the components directory, the header, the footer, updated the layout, and gave us a summary of all of the different changes. Again, we can see all of those different edits there. Now that I've applied all of those different changes, I can take a look and save these out one by one. I'll save everything out and I'll take another look at our application. So now we see we have our header, we have our footer. Then as soon as we have the changes that we want, we can go ahead and we can update our GitHub repository and push all of those different changes. Here we see we created these net new files and then we also had some files that we changed and we can go ahead and push that up and we can have that reflected. The really nice thing with this is you can see how easy you can spin up all of these different remote agents. And one of the big benefits of this is instead of having a graveyard of the quote we'll get to it tickets, all the different brittle tests, stale docs, tiny bugs, or those pesky little paper cuts that add up within your codebase over time that ultimately never really climb the sprint wall. With remote agent, it allows you to easily handle those smaller tasks and hand off to an agent with clear instructions, which will ultimately help you clear the backlog and ultimately allow you as well as the engineers that you work with to not have to worry about some of those smaller tasks that an AI agent could ultimately do. In terms of some other use cases, there are a ton of different ways on how you could potentially leverage this. whether it's spinning off the agent to help you update the documentation or to refactor different portions of your code or even just have an agent that will write unit tests for you. So while an AI coding agent won't be able to do everything for you if you give it very poignant instructions, it will be able to eliminate a lot of that engineering toil that you would have otherwise had to spend on all of these little tasks deprioritize if you didn't have access to an agent that can run in parallel. Now, in terms of some best practices for the agent, you can effectively think of this as an army of eager interns. So, you can spin off all of these different bite-sized and self-contained tickets that you want to have the agent perform. Now, some other helpful steps is if you actually include instructions on how to run your test suite or llinter or all of the different checks that you have within your codebase, you can instruct the agent to perform that function. Again, the agent does have access to your terminal. If you do have different tests that you do want to validate, whether it's unit tests or integration tests or whatever it might be, you can just include that within the instructions for the agent to run. Now, it goes without saying, make sure everything that is merged within the remote agent you review, just like you would a PR. You should not yet trust the agent, not quite yet. While it does do well at some tasks, you will also want to review all of that code that was otherwise AI generated. Otherwise, that's pretty much it for this video. If you found this video useful, please comment, share, and subscribe. Otherwise, until the next