Use Slack as an OS with GitHub and Cursor Agents (Full Transcript)

[00:00:00] Speaker 1: Do you want to see what it really means to use Slack as an operating system? Watch this. Hello and welcome back to Slack School. My name is Mike Reynolds. I'm your host. I'm part of the Slack team here at Salesforce, and today we're going to be starting a new miniseries where we look at Slack as an operating system. I want to spend just a moment and talk about this idea of an operating system and what do we really mean when we say Slack is an operating system? Well, I like to think about it like the way I think about my phone. If I need to get something done, I install an app on my phone and then I'm able to use that app to get my job done. Slack is basically the same thing. Slack is a central tool that you install your apps in, and then no matter what you're doing, you can use Slack to get that work done. We're going to dive into this new series with my buddy, Matt Justice. He's going to be taking us through several different examples. Today we're going to focus on a new one. I'm excited to see what he's got because we're going to get to see getting work done in Slack using a different tool. Let's get after it. Hey, Matt, how's it going? Great, Mike. How are you? I'm doing fantastic. Why don't you just introduce yourself to first time on Slack School?

[00:01:22] Speaker 2: Hey, everybody. I'm Matt Justice. I'm a part of the Slack team here at Salesforce.

[00:01:27] Speaker 1: Fantastic. All right. What are you going to show us today?

[00:01:30] Speaker 2: We're going to talk about a very cool integration from a company, Cursor. They have a coding agent that lives inside Slack and can actually do a lot of really cool things for our developer friends out there. What we're looking at here, we have a payment bills channel. This channel has anything that is related to a web application that receives payments. Think of this as you go to a web page, you need to fill out your payment information to pay for a product and ship it out. Now, you can see here that we also have a GitHub action results. This is based on a test that GitHub runs when we create a new branch, but I'm not really sure what's going on here. Before I get into the weeds, I'm going to ask Slackbot a couple of questions here. Can you tell me about the recent failures of our web payment app in this channel? Now Slackbot will look at the history of these recent failures and give us a little summary of what's happening based on what other coding agents or other applications or what folks have been talking about here. So here we see a quick summary of the recent failures that are happening. It also has noticed that since this is a demo environment that we keep having the same issue over and over because we're merging and reverting a branch. So Slackbot's pretty smart. And the other thing that I'd like to ask Slackbot about here is, are we seeing this in any other channels? So I could ask, you know, is there a more concise way that you'd want to see this information? But I kind of know what's going on. So I just want to ask, are there any other channels where this is being discussed? So this would give me some context on, you know, are there other feedback channels? Are there other areas in the business where we might be seeing a similar issue to give me context on what's going on? So the other thing I'm curious about is, is there any business impact or any other channels where this is being discussed? So I'm just going to ask Slackbot, are there any other channels where this issue is being discussed? And is there any business impact? Now Slackbot should should look at all of the channels that I have access to and provide me some information based on, you know, are we getting feedback from other channels? Perfect. So we can see that this is obviously being discussed in the channel that we're in. We're seeing some impact around actually the potential monthly revenue being lost of $180,000. So hey, this is kind of a big deal. So we want to see what we can do to get this fixed as quickly as we can. Now my day job is not coding, but with the tools that we have in Slack and with Cursor, we should be able to figure this out. So we have some details in this thread here of where the failure is happening. You can see this is the repo that it's associated with. This is the branch and this test has failed. So we're just going to come in and reply to this thread and we're going to invoke our Cursor agent and say, Cursor, can you investigate this? So the Cursor agent, we added into this workspace and we've done some configuration on the back end to allow it to look at this particular GitHub repository, allow it to read and write. So here we can see that Cursor has launched a cloud agent and it's going to investigate this particular repo. So it's identified from the context of this message that this is the repo that it should go look at. We have given permission within the Cursor environment and it's going to use this model to go and look. I could move on with my day, continue going to meetings and let this agent work, but I kind of want to see what's going on under the hood. So if we toggle out to Cursor, this is what's actually happening. So the Cursor is setting up an environment, it's going to look into the specific issue. It'll try to identify a root cause based on the test and based on the recent changes that have happened, and then it will either propose a fix and or create a pull request for us based on the investigation that it does. Now again, this is a non-deterministic agent, so it may do other things as well. Sometimes it does screen recordings, which are very cool. So you can actually see what the new code would look like when you're actually using it in production and then we can decide whether or not we want to move forward with merging that branch based on the changes it's proposing. Now the other thing that's very cool is if you look here on the left-hand side in Cursor, you can actually see all of the other agents that have been running. So this happens to be one issue that we're looking into and this probably will take a few minutes, but we could also have a series of agents looking at more complicated issues that could take days, hours, weeks to go and investigate. So if we're looking at something really complicated, we can set an agent off and have it do something much more complex.

[00:06:52] Speaker 1: And so when you're doing that, if you have these really long running tasks, you know, I want you to evaluate a massive code base, it's just going to take some time that when that agent comes back, that still happens right in Slack. So you've sent that first message and then maybe you don't get a message back from the agent until tomorrow, but it's all native in Slack?

[00:07:11] Speaker 2: It is all native in Slack. And now you'll see here, the agent actually updates the message that it has to you within Slack. So it depends on the type of coding agent that you're typically using. But with Cursor, it's about every 30 seconds, this message will update, giving us an update on what is actually happening. You can also see that Cursor has kicked off an emoji, so a checkmark of, hey, it's looking into the issue. So it found some issues associated with this, these recent changes. And we're going to ask again, can you create a pull request to resolve this? Now, again, you can see that the Cursor agent has started up again. If we toggle back to the environment within Cursor, it's starting and planning its next moves.

[00:08:02] Speaker 1: So you can install the GitHub app right into Slack, and then you've got your Cursor agent right inside Slack. And so you've actually got all three of those tools, Slack, GitHub, and Cursor working together.

[00:08:12] Speaker 2: That's correct. And what we're seeing here now, so the Cursor agent has now investigated that issue based on the GitHub repo. It's identified that there is a test negative payment amount that is expecting a positive value, but it's allowing for negative values to come through. So long story short, we're allowing for customers to pay negative money, and we still ship out the product, which is why we're losing around $180,000 a month. Now, you can see here within Cursor, it's telling us what those changes need to be. If we look back in Slack, we're seeing those updates happen as well. Now, you can see that it's also creating a pull request. So if we want to go out and look at that pull request and see what Cursor is proposing that we change, again, based on the context that it has, it's giving a summary. It's going through a series of tests, and it's actually going to show us that this positive value is what we actually need to resolve, ensuring that we're actually collecting money rather than negative money. So when this is finished, we'll see a summary here, and I could go in and hit Ready for Review and merge my branch into production. Now where this is also very cool, because we have this channel also connected to GitHub, you can see that this initial test of that branch has succeeded, so we're not seeing that failure any longer. And here in just a moment, it does another check on the main branch just to make sure that we didn't break anything, and we can see that our Cursor pull request that was created has resolved our problem, and we should be in a happy place. And there we have it.

[00:10:17] Speaker 1: You know, I could see this being really, really helpful at the beginning of a sprint when you have a, you know, oftentimes for me, I would have a spike ticket where it was like, just go investigate how plausible something is, or, you know, is this approach reasonable? Is there a better approach? And then being able to just stick an agent on it to say, hey, I want you to start this, and maybe I have, you know, an entire day's worth of points against that one story, and then that Cursor agent does it, and then I come back working on something else, and then by noon, you know, I'm an entire day ahead because Cursor just took care of it. That's awesome.

[00:10:54] Speaker 2: One of the really cool things that we've seen with coding agents in Slack is that it allows our engineers to focus on, you know, more strategic work. So if there's a new code base that comes out from AWS, they don't have to go in and investigate how exactly that works. They can actually use coding agents to say, this is what we want to go and do. Can you please investigate, you know, this new code base or this new integration pattern? And how would I be able to go and do that? So we can focus more on solving customer issues rather than, you know, the exact syntax of writing code. Thanks for having me on today, Mike. I just want to give a quick summary. So this Cursor agent went in, identified that we need to fix our code to reject non-positive payment amounts. You can see the summary here on the right. It created a pull request for us that we were very quickly able to merge into production. We had another test in GitHub that was validated here so that we know everything is working properly. Now, this is a very cool use case of one coding agent working with an app with GitHub in Slack, but we have a lot more. So one area that I think everyone should check out is if you come inside of Slack and come out to our agents tab, we have the agent exchange. So here you'll see Cursor, which we have installed, but there are a ton of other agents that you would be able to very quickly discover and go in and add with the click of a button. So go out and check those out. Thank you again for having me on and I'll see you next time.

[00:12:22] Speaker 1: Well there you have it. Using Slack as your operating system to bring in tools like GitHub and Cursor is easy. Let us know what you thought in the Slack community at slackcommunity.com. Join us in the Slack School channel and give us some feedback about the episode. Let us know what you want to see next or just say hi. We'll see you next time. Hey, you did a great job today. I can't give you attention right now. I have to do my job. You're not supposed to stick your butt at the camera.