Use Aisle Tasks in ChatGPT, Claude and more

Hi, I'm Alex, founder of Aisle, and today I wanted to take you on a quick preview of how to take the Aisle tasks that you built and use them inside of Claude, ChatGPT, Cursor, Claude Code, Codex, it doesn't matter. Something we're rolling out to all users early next week and we've been thinking about a lot. If you've been watching our content, our videos, reading our blogs already, you know the way we think about Aisle is this the best place to build these deterministic AI mini applications that you need to start building your business on. And very frequently, you want to orchestrate those through code and use multiple models so you can manage token burn, have easier upgrades, and trust the orchestration of the task. And today, what a lot of our users are doing is taking those tasks and are engaging with them through an Aisle chat or project, outputting them to Slack, having it send them an email, create a report in Google Drive.

But where we really succeed and help our users is as the best place to create those AI-driven applications. We don't mind where you put them and happy if you want to put them in Claude Code or, you know, Claude or ChatGPT or anything like that. So, I built this very basic task here. I'm sure you will build something more complex for your business. All this does is takes the user's input and again, it's quick task preview.

Runs this prompt that tells a pirate joke based on the topic that the user's provided. It's not a very long prompt and puts the output here, but it's enough to start talking about that execution server. So, once we have that, I'm going to go in and press developers and come to this execution MCP server. I'm going to create a new one here. I'm going to pick that pirate joke generator and I'm just going to call this demo video.

I could assign as many tasks as I want to this server. Once this has been created, you'll notice there are instructions on how to add this for all of the different platforms here. Before I go and do that, you might actually be wondering how do you give this to other people in your organization? Well, you have a few easy choices. One of the things you can do, consistent with everything else in our platform, is come in, press sharing, and either share this org wide or start typing in specific people who have access to this server.

But, once this has been created for Claude, what I'm going to do is copy this URL that's been generated here. Then I'm going to come into Claude, and I could add this in the connectors directory, I could add this in projects. I'm just going to add this here because it's a little bit easier and faster, and I'm going to add a custom connector. Now, depending on your organization's Claude settings, you might need to enable end users to be able to add custom connectors or have an admin add it for you. Depends on how your organization is being set up.

Here, I'm going to call this demo video. All right, let's call it pirate joke, so it's a little bit more specific, and it turns out that even on a demo video, I'm going to make typos all over the place, and let's put in the URL. And that's been added. It's going to go through and ask us to authenticate here, and I'm just going to use this demo account I made for the call here. And this is going to go through and basically authenticate us into the platform so that we can use that connector that's being built, right?

It's not like the URL is just going to work without authentication, that's an OAuth. So, what I can do here now is I can say, you know, "Can you please use the demo video pirate joke connector to tell a joke about a C PHP. I'm on a demo video right now." All right, so we don't want to ask it too many questions. There are many ways you might do this, but I just want to show you that the actual task is going to run in this environment. You'll notice it's asking us, I'm going to say that it's always allowed to run this task that we built here.

And this is great, right? I'm able to come in. It's running this thing that we built in another environment. We control it, we can represent it in code, we can start interchanging the models as much as we want. And you'll see we've got this response here, and within Claude, I can actually come in and expand and see what was it, what got responded, how did it all work.

So, really good setup in my mind. And I can actually come back here and see that it was and see what the trigger was and who the user was in case I want auditing or logging on what's actually happened with these I'm distributing. So, really excited to start rolling this out to all of our users. Uh something we've been thinking about planning for a while and you'll be able to start taking these deterministic AI orchestrations, tasks, processes that you're building and use them wherever you would like.