Ever wished you could build your own AI agent without writing a single line of code? Well, in under 20 seconds, I just made my own financial assistant that performs rag over a knowledge base. Now I'm chatting with it. I can see the response streaming to me in real time and I can see the tools called as the agent generates its response. Introducing Open Agent Platform by Langchain. the citizen developer platform for creating, managing, and deploying intelligent land graph agents in minutes. In this video, I'll be covering how you can start using our managed version of open agent platform, including an overview of the platform and all of its key features, configuring your own document collection, which will be important for enabling rag in your agents, out of the box tool support powered by Arcade, immediately giving your agent access to hundreds of different tools, and finally adding and interacting with your own tool and supervisor agent. These are what we support out of the box in this hosted version of open agent platform. In another video, I will demonstrate how you can configure everything that you will need to deploy your own version of open agent platform to production. It will include adding your own custom agents, plugging in your own tools via your own MCP server, setting up your own rag server via LANC connect, and configuring enterprisegrade authentication. Now, let's dive into it. Now, we're on the chat page. This is where you interact with your agents. Simply select an agent from the drop down like the default assistant and start chatting. It's perfect for testing and refining your agents behavior in real time. Now we're on the agents page. Here you can view and manage all of the agents that you've created. In the hosted version, you'll see the tool agent and the supervisor agent. This section lets you configure their settings. The tools section lists all the external tools your agents can access. These are powered by the integrated MCP server, giving your agents the ability to perform tasks like web searches, data lookups, and more. You can browse the available tools and explore their capabilities right from this menu. And finally, the rag page. Rag stands for retrieval augmented generation. In this tab, you can manage your document collections. essentially the knowledge bases your agents can search over. You'll be able to create new collections and upload documents, making it easy to give your agents custom knowledge. And that's the basic layout. Each section is designed to be intuitive and accessible, so you can build, manage, and interact with your agents without any coding required. Our hosted version of open agent platform comes with support for hundreds of tools out of the box. They leverage an MCP server provided by Arcade. In this tools page, you get an overview of all of the tools that are available to call from your Langraph agent. For each tool, you can preview its details. This will show you its schema as well as a description about how that tool performs. Additionally, you can enter the tool playground, which will bring you into a new page that will allow you to test out that tool's functionality. As an example, I've opened up the math tool, which simply adds the other two numbers. I can specify the values of those numbers to add and then I can run the tool and see the output rendered on the right side. If I want to stay in this window and test out the functionality of a variety of different tools, I have the option of a drop down that allows me to select from the entire list available to me. For tools that require any kind of authentication, once I configure it and I try and run it, I'll be presented with a link that will generate a call back to go and provide the necessary authentication for it. I'm going to go ahead and authenticate my email and then navigate back to the open agent platform. Then when I try and run the tool again, you'll see that the response will be successfully populated. Let's go ahead and navigate over to the rag section in the left menu bar. This is where we will configure our collections. A collection is a repository of documents that are exposed via a rag API server called LANC connect that is callable as a tool for your agents. In this hosted instance of open agent platform, we've implemented LAN connect for you. But if you were self-hosting, then you'd need to manage this yourself. Let's go ahead and make a new connection. We'll call it financial statements collection, which will be a place for us to store some of our financial documents that we intend to use for RAG. Once we're on this collections page, you can see over in the left the name of the current collection that we're in, as well as the ability to create new collections. You'll also see on the right where we add documents to our collection. We can do this by either dragging documents or from simply navigating through Finder. I'm going to go ahead and drag in Google's most recent 10K report from 2024. Also, I want to add a text snippet to this collection. Great. Now that you finished adding documents to this collection, it's ready to go for use in your rag agents. Optionally, you can choose to chat with your documents to see if they are being retrieved as you expect. I'll save that for when we begin to build with our agents. Great. Now that we've created a financial documents collection where we uploaded Google's 10K report, let's go ahead and create a finance rag agent. The sole purpose of this agent will be to perform rag over our finance documents collection. Next, I want to create a Google Sheets agent. This agent will be responsible for creating Google Sheets upon request. Finally, let's create a supervisor agent that will orchestrate workflows between our recently created Google Sheets agent and financial rag agent. I'm going to go ahead and modify the system prompt as well, just to give the finance supervisor the extra bit of context that it needs to actively orchestrate its workflows. First, let's get Google's 2024 topline financial results according to its 10K. I want these to be returned into a table format so that I can leverage my Google Sheets agent to create a spreadsheet from the result. Awesome. Now that I have these results in table format, I'm going to ask the Google Sheets agent to make me a new spreadsheet. You can see here that because this is an MCP tool that requires authentication, I'm required to authenticate with my Google account. I'm going to go ahead and do that now. Now that I've provided authentication, I'm going to tell my supervisor agent to try again. Awesome. And we can see here that it successfully created the spreadsheet with Google's topline financial results from 2024. We've just seen how easy it is to create a supervisor agent that orchestrates individual specialist agents. In this case, a financial rag agent and a Google spreadsheets agent. That's the power of open agent platform by Langchain. In under five minutes, I created three separate agents, two specialist tools agents and one supervisor agent to orchestrate a workflow with them together. With our self-hosted out-of-the-box offering of Open Agent Platform, you too can create complex workflows. And if you're feeling like you want even more customization, then make sure to check out our video on how you can do this yourself to get everything that you will need to deploy your own version of Open Agent Platform to production. In the meantime, happy building.