How I Run 6 Coding Agents at Once (My Actual Workflow)

I've already done tutorials on how I use Cloud Code, how I use Cursor, but I wanted to do a video to show you guys what my actual workflow looks like when I'm in the weeds. How I'm juggling everything, how I'm prompting, and how I'm moving quickly. This one's a little bit more advanced. It is not a beginner tutorial. If you want a more beginner-friendly tutorial on Cloud Code and Cursor, I will leave a link in the description to those videos. If you're new here, my name is Chris and I build productivity apps. And as a solo developer with multiple apps, I am constantly running like five instances of Cloud Code at once. I'm on the $200 a month max plan. even running it the way that i am against multiple code bases for many many hours each day i have not run into any limits but in this video i want to show you realistically what it looks like to run this many instances at once my workflow has changed a lot because of ai coding and honestly i feel like ai coding in general has really changed over the last year it used to be about typing the right commands and telling the agent what to do tab complete was a thing like that used to be what ai coding was but now i haven't even done that in the last year now it's all about having conversations with agents like cloud code the way that i work is I genuinely treat the agent like a human. I'm constantly asking questions like, what do you think about this? Is there a better way to do this? And I'm having this kind of back and forth every single day. Let me just show you guys what I mean. So let's say I'm working in the Luna code base, which is my budgeting app. I'm having a conversation with one agent and I'm asking it, is there any place where I'm using SwiftUI gestures where it might be better to use UIKit? And then I let that run. And when you're using Cloud Code like this, it is really slow because it's thinking a lot. It takes like three to four minutes to respond sometimes. And since I have so much time, that's when I switch to another agent. By the way, as you saw, I'm using dictation to work with Cloud Code, and that is super important to my workflow. The tool that I'm using is Whisperflow for dictation. More on that later in the video. But because I'm dictating everything, I'm able to get way more detailed prompts compared to if I was typing. And I'm able to move very quickly between agents. So now I've switched over, and now we're looking at the code base for Amy, which is my calorie tracking app and I'm working on another feature which is barcode scanning. So now I'm gonna ask it, I'm thinking about adding barcode scanning, can you do some research and figure out if there are any good APIs for this? What's the best way to do this? And make sure that it goes well with the current photo scanning that we have. And then I kick off that agent while the first one is running. And I have a bunch of other apps too. I'm working on Ellie, which is my daily planning app. And I do a lot of AI consulting work. So I build a lot of prototypes. It's just a bunch of code bases that I'm working on at once. Also, if you didn't notice, I'm using plan mode in Cloud Code. The way it works is you hit shift tab and then you enter this plan mode. This is a mode where Cloud Code is not going to start editing files. It's going to draft a plan on what it's going to do. And then I can go back and forth with the AI to really try to reshape this plan before I have it execute. I never skip this. step and then when i'm happy with the plan that's when i let it go off and execute another tip is you can run this command with clod code which is dangerously skip permissions and if you've ever worked with cloud code you know that it's always asking for permission can i update this file can i go look at this website when you run this command it's going to skip all that and it's just going to go off and execute without asking your permission for anything it is a little risky to run this but i am personally doing this if i'm working in an area of the code base where i feel like not a lot can go wrong And to be honest, I think as long as you're using plan mode and you are really reviewing the plan that it's generating, doing that along with reviewing the code after it's generated should be completely fine. So sometimes I'm running this dangerously skipped permissions, especially when I'm running like five agents at once, but I'm always using plan mode and I'm reviewing the plan every single time. And I'm not just using these agents across multiple code bases. I'm actually running multiple agents in a single code base sometimes. So for example, in Ellie, my daily planning app, I really want to improve this AI assistant feature that we have. And so I'm going to ask it, Can we actually improve the animation when the loading happens? I don't want a circle, but I want a shimmering animation where it kind of shimmers from left to right. And then at the same time, I have another agent in the same code base. So it's just another Cloud Code tab. And I'm asking it about the Google Calendar integration. Are we missing any actions that Google Calendar supports? Can you look at the API documentation and see if there's anything glaring we should be adding? And then I'm going to kick off a third agent also in the LE code base. And this one, I'm going to do the same kind of research I'm doing for Google Calendar, but with Apple Calendar this time. I wrote this integration a long time ago. I don't even know if I did it the right way. So I'm going to ask it, is the method that we're doing for Apple Calendar integration correct? Or is there a better way we should be doing this? And then I'll actually kick off a fourth agent to check the code quality of one of the features I was working on earlier. So I'll say code quality check, please. By the way, that was using a Whisperflow snippet, which I'll talk about later in the video. So I'm just kicking off all these agents and going back and forth with them. Because of how long it takes, there's a lot of downtime. And so while they're thinking, I can just hop between each of them. And this is how I'm moving so quickly. We have six agents running simultaneously right now. One more tip, if you're in Cursor and you have a bunch of terminal tabs open like this to run Cloud Code, you can actually rename these terminal tabs, which I do to make sure that I know, okay, this terminal tab is for this conversation. We're working on this feature here. So let me just go ahead and rename these. Now I renamed it. This helps a lot when you're juggling multiple agents. And then as you probably saw, the key here is that this would not be possible if I was not dictating everything. As the coding models are getting better, it's less about prompting in a specific way or saying the right set of words. It's more about giving the right context and intent, which is way easier with dictation because then you can just word dump all the context, just like I would to a human if I was speaking to them. And then the model will understand and perform the task better. I get a lot of questions asking what I'm using for dictation. And the answer is whisper flow and you're saying to them for actually being a channel sponsor whisper flow is a smart speech to text app It works in any application. So if using cloud code cursor slack superhuman or any app with a text box It will work there and by the way the thing that says dictation using whisper flow It's something I enabled for the video if you're using it That's not gonna appear there for you They also have integrations with cursor windsurf and cloud code So it can actually understand your code base and you can tag files and variables without typing anything to check this out if I say something like, can you check for code quality in subscription overview.tsx? It correctly tags the file without me having to do anything. They also have an iOS app, which I'm using all the time. When I'm responding to YouTube comments, that is typically how I'm doing it because I want to be really thorough and it's just so much easier with dictation. And then obviously I'm using it in iMessage and Slack too. But dictating is so important, especially when you're running multiple agents, because you can get... way more detailed prompts than you would otherwise if you were just typing everything. Let me show you another example of what I'm talking about. I'm actually going to ask the agent, can you do some research on MCP servers? I'm really interested, especially on external MCP servers versus internal, because I know that there are slight differences in architecture. That would have been super annoying to type out. Like I probably would have just stopped at, can you do some research on MCP servers? And that's it. But I can get way more granular because I'm dictating. And the last thing I want to show you is their snippets feature. You can save snippets for frequently used text. And that's how I did that code quality check snippet for Ellie earlier. So I have it where if I say code quality check, please. it expands to this prompt that i have saved which saves me a ton of time because i'm constantly using this to check code quality i will leave a link in the description if you want to check out whisperflow there's a code for one month free and that's on top of the 14-day trial you already get with whisperflow pro they did not have to give me the code i'm the one that asked for that so a huge thank you to them for providing that so that's my workflow this is how i am juggling multiple cloud code instances and again i have not hit any limits being on the 200 month plan and I'll leave you actually with a little tip that I figured out. If you've used Cloud Code, you know that there are weekly limits. I don't know if this is a glitch because I really haven't heard anyone talk about this, but if you choose to upgrade to the yearly plan of Cloud Code, I don't think they have weekly limits because that's just completely gone from my UI. So if you're someone that is running into the weekly limits, even on the $200 plan, which that's really rare, people normally don't hit that limit, you could upgrade to the yearly and... those limits are gone. The important thing for my workflow is always use plan mode. You can use dangerously skip permissions if you want. I think it's completely fine, but again, use plan mode and then dictate everything if you're doing this. If you've been following along, you know that I do use cursor for some things, especially complex debugging, because I love their debug mode. If you want to see a video where I talk about that, I will leave a link in the description for that as well. I have more in-depth guides to my workflow that I'll leave below with tips like MCP servers, how to use sub agents, more granular cloud code tips like that. I'll leave a link in the description. If you enjoyed this content, feel free to check out my Instagram and TikTok. I post almost every other day about building productivity apps. And obviously, if you like this content, don't forget to subscribe. But thank you guys so much for watching, and I will see you guys in the next video.

Hi my name is Chris and I build productivity apps 👋 

---
Dictation tool I'm using (Wisprflow): https://ref.wisprflow.ai/chris-raroque
Use code RAROQUE for 1 month free 👀
---

My apps and socials: https://chrisraroque.com
My agency (work with me): https://aloa.co/

Timestamps:
0:00 – Intro and what we are covering
1:09 - Working with multiple agent (and plan mode)
2:42 – Dangerously skip permissions
3:26 - Multiple agents in the same codebase
5:04 – Dictation & Wispr Flow (My favorite dictation tool)
7:28– Claude code max limit tip
7:59– Summary and final thoughts :)

#appdevelopment #dayinthelife #softwareengineer #startup #softwaredev #indieappdeveloper #dayinthelifecoding #codewithme #buildinpublic #vlog