GLM-4.7 + Conductor: This how I'm running more than 100 SUPER AGENTS AT ONCE! (for low cost)

Hi, welcome to another video. So, I have been loving working with GLM4.7. It's very much like Sonnet at a cheaper price and faster speeds. I feel that it's much better than Sonnet, but a bit worse than Opus, which is completely fine as it is about 20 times cheaper than something like Opus or even Sonnet for that matter. I have been searching for a background agent for tasks that doesn't burn my wallet. I have been finding that GLM4.7 is currently the best for that. I can

make it create a plan and then talk with it about the plan and then make it implement that which is quite great. I find it really good for that and for almost all kinds of workflows. Now, I have been using it a bit differently than I generally do. And with these new models, I am leaning more towards letting them do all the work and then just reviewing the code when I make the final PR. And that's why I want to have a setup that is less cumbersome to open and assign tasks to, but still

keeps me in control. And I have figured out quite a good workflow for it. And I wanted to share that as well. So I have been using Verdant a lot these days for these tasks. I think currently their cost is quite good. The models are not nerfed unlike many other places and it gives me all kinds of features like git work trees and whatnot and you also get good code reviews with it. It is one of the only two subscriptions that I have currently. It doesn't have GLM4.7

or Minimax yet. And that's fine because I want to keep it as my main coder, but I'm also using GLM4.7 a lot these days because of its insanely lowcost coding plan. I mean, I can pay for a year's worth of GLM coding plan max subscription for just $280, while the same plan for something like Claude will cost way more. I think Claude's plan and value is really great, but in this case where I already have the verdant plan, it doesn't make sense for me to have that in order to have a background agent

through which I can let it run on long tasks as it can run out of the usage limit quite fast. That's why I'm really loving the GLM 4.7 coding plan. However, I really wanted to be able to use something like Verdant that works with GLM 4.7's coding plan. And after heavy consideration, I have selected Conductor. Conductor is basically a cloud code and codeex key. It is Mac OS only, which is a bummer if you use Windows. I'd recommend you to check out Claudia, which is now apparently called Opode, or

just search for my videos about Claude Code, guys, and select one. There's a lot of them. There's also Conduit, which is also quite great, and it's crossplatform. I find Conductor to be the best because it has the features that I need. It is quite optimized for Mac OS and there are some more cool features there which I'll show. One of the major things is that whatever you configure in your cloud code settings, it uses that exactly or you can also set the environment variables

just for conductor if you want to do that. So let's go step by step and let me show you how I'm using it all. First, you got to go to GLM and then get your API key from there. I'd recommend you to just get the $3 plan and check it out and then upgrade from there. So, you can check that out. Anyway, then you can go ahead and set up your Claude code config as per the docs. It's quite easy to do and you can refer to my previous videos as well if you want a full config.

If you set it up directly in claude code, then it will automatically work in conductor as well, despite whatever model you choose there. Make sure that you also set up the default sonnet and opus model in order to make sure that everything works correctly. If you want to keep your claw code as is, but want to use conductor with just GLM4.7, then you can do that as well. So, you just have to install Conductor and then open it up and then head on over to settings and set up the environment

variable with the GLM 4.7 environment variables and you'll be good to go. You can use whatever model here and it will redirect you to GLM 4.7. Remember that in the system prompt it tells it that it's Sonnet. So, if you ask it then it will tell you that it's Sonnet. If you wanted to say that it's GLM4.7, then you'd have to make sure that you set up the defaults on it and Opus model in the config as GLM4.7 and then it will say that it's GLM4.7. But if you don't care about that, then

just the environment variable setting works fine. Anyway, now the major thing that conductor does is that it is super focused on git work trees. I have talked about git work trees before, but if you don't know, then it's basically clones or branches of your projects that allow you to make sure that whatever you do, you have the older state available to you. Once you are done with a task, you can create a PR and then get that merged. It is really good in my

professional workflows as I like to have things streamlined. The reason I use Verdant a lot is because of their amazing git work trees feature as well. And conductor here just doubles down on that because it probably doesn't even allow you to edit the project files directly. It always makes a branch and then you merge that back in with a PR or directly. So I like this a lot. It creates branches with city names by default. What I do is that in whatever project that I'm working on, I create a new

branch and do a good chunk of changes and then use the create PR option in conductor itself and then run code review on them and get them merged. I do plan to make a video about my whole workflow about how I maybe ship a feature in a project end to end. So let me know if you guys would want a video on that. Anyway, in one branch, you can make multiple threads and everything. When I'm doing something, I first go to a new thread and then just select whatever here. All of them just point to GLM4.7.

I select sonnet for safety. Anyway, then I select the plan mode. I try to make a plan with it about whatever it is that I want to implement. go back and forth with it. And conductor actually prompts it to ask for follow-ups, which works quite well with GLM. I would have liked if I could set up like a different planning model in here. I think that could be done. If someone knows how to do that, then please comment below and help me out. Anyway, after the plan is made, I get it

implemented. And conductor also keeps everything in auto approval which is great and it maintains git and everything for me which is quite a lot of stuff. I don't like to look at code for each change. I only keep an overall look of the plan to understand the major architecture changes that it's doing and if it's efficient or not. In the PRs I check for the code quality and then get that fixed. I can run multiple feature threads in multiple workspaces with it and work with Verdant

implemented. And conductor also keeps everything in auto approval which is great and it maintains git and everything for me which is quite a lot of stuff. I don't like to look at code for each change. I only keep an overall look of the plan to understand the major architecture changes that it's doing and if it's efficient or not. In the PRs I check for the code quality and then get that fixed. I can run multiple feature threads in multiple workspaces with it and work with Verdant

on super complex issues if I want and it is quite good. It takes up less memory than using whole VS code and VS code is not very focused which is not very cool. So this is the best config that I was able to do and have been using basically for all my code. It helps me save a ton of money. It's quite good and works very well. That is majorly about it. Let me know if you guys would want a full workflow video with code review and everything as well. Overall, it's pretty

cool. Anyway, share your thoughts below and subscribe to the channel. You can also donate via Superthanks option or join the channel as well and get some perks. I'll see you in the next video. Bye.

cool. Anyway, share your thoughts below and subscribe to the channel. You can also donate via Superthanks option or join the channel as well and get some perks. I'll see you in the next video. Bye.