Create MCP-Enabled AI Teammates with Runbear in 10 Minutes

--- type: transcript date: 2025-05-05 youtube_id: iHV8xCG-7LU --- # Transcript: Create MCP-Enabled AI Teammates with Runbear in 10 Minutes In this video, I'm going to show you how you can create your own AI teammates using RunBar. Over the past couple months, you've likely seen MCP catch fire with their already being thousands of integrations available for most services you can imagine. Runar lets you tap into this ecosystem by seamlessly integrating the services you're interested in using directly in the platforms you and your team are already using. And with Smart Reply, there's no need to mention a chatbot or spend time searching for info. It will provide relevant information exactly when you need it without any additional friction. Let's see how you can set this up in under 10 minutes with no coding required. So now just to quickly demonstrate how this works. What you'll be able to do is you'll be able to add multiple runar assistants within something like Slack. What I have within here is an AI teammate that I called research helper. And for this research helper AI assistant, what I gave it access to is the firecrawl MCP server. What that MCP server allows me to do is research particular topics or go and retrieve different URLs. What I can see within here is by just calling on my research helper, what it was able to do is to gather all of this information. I asked for the latest research on GPD4.1 for a technical audience. And we can see that it responded with this really detailed and technical overview on all of the different details. We have model variance, context window, knowledge cutoff, output token, basically everything that you would need to know. me as a technical content creator, where this could be helpful is exposing it to things like my editor or potential companies that I work with. Now, the one thing to know with RunBear and the assistants that you set up is they're really all going to be geared towards whatever specific use case you're looking for. Now, just to show you how easy it is to get started, if I head on over to the assistance tab, we can click add assistance and for MCP enabled servers, we can click the cloud implementation here. In terms of setting this up, all you have to do is determine the name. The name can be whatever you'd like. Then from there, we can determine what the system prompt is. This is largely going to be how the actual assistant behaves. If you want it to follow particular directions, you can go ahead and instruct all of that here. There's also a handful of examples. Say if it's a Q&A bot or a contract review. There's a number of different options that you can choose within here. And when you choose the template, it will just swap it out. Within here, we can put in as much detail as we'd like. We can also put in a few different examples. That's one thing that a lot of AI models will do well on is if they have some examples to follow. It will give an idea in terms of how to respond with the system prompt. You could start simple and just start to add on it over time as you start to see how the assistant begins responding within this. I could say you're a research assistant. You'll take the queries that I ask and search for the relevant information and perform a form of deep research on the particular topics that I'm asking for. One thing to note is you will have access to the latest anthropic models here. Everything from cloud 3.7, cla 3.5 all the way through to opus as well as haik coup. So now at time of recording they do have a complimentary service for anthropic to use. Alternatively, you can always bring your own API key. You're going to be build directly at the rate of whatever anthropic would charge in terms of the amount of usage that you use for the model. In addition to setting up an MCP server like I'll show you here, you can also add in integrations like Google Drive, Confluence, Notion as well as Slack. Or alternatively, you can upload your own documents. So things like PDFs or CSVs or what have you. The nice thing with the MCP servers within RunBear is they have a marketplace directly integrated within the platform. Now, alternatively, you can go and add in your own MCP server by declaring the name and also adding in the URL. But in this case, I'm going to show you the marketplace. So, you can go through the list. You'll see a ton of different examples. Basically, just about everything that you've probably seen in terms of the different MCP servers that are out there, you'll be able to find within this list. In this example, I'll show you how you can use something like Firecrawl. If it's an MCP server that does require an API key, you'll have the option to choose between whether each user will have to provide their credentials or alternatively the shared credentials. So, within here, I'm going to select the authorization method to be shared. Now, the one thing to note with shared credentials is you will just have to make sure that you do trust the users with your credentials because depending on the service, this could have access to potentially sensitive information. That's one thing just to be mindful of. But you can also specify per user as well if you'd like. Depending on the MCP servers that you select, you'll be able to see all of the individual tools within those servers within here. I can scrape pages, extract web data, so on and so forth. I'm going to go ahead and connect and add server here. And then lastly, the one thing that you'll have to implement that's going to be unique depending on the MCP server is the API keys. Within RunBar, they leverage a service called Pipe. And this is going to be how you can securely leverage those different services. I'm going to go ahead and click continue. Now, I have the option to go and create this. Now, once it's created, what we can do is we can select where we want the assistant to connect to. So, in this instance, I'm going to leverage Slack, but just know that you can use this within Discord, Microsoft Teams, HubSpot, or Zenes. Once you select the platform, now if it's the first time that you're integrating within a Slack group is you'll have to go through the steps of just getting the authorization token from Slack. It will walk you through very clearly exactly how to do that. And then once that's set up, you'll see all of the different configuration tokens within here. I have a number of different Slack channels. From there, I can go ahead and click create. From there, you'll see Researcher 2.0 is requesting permission to access the example developers digest Slack workplace. From there we have some configuration steps that we can take. We have the application within Slack. What we can do is if I make a new channel, I'll call it Run Bear demo. And then for the first time that you use the chatbot, what you can do is you can at mention the chatbot. I can say something like, "Tell me all of the information that I need to know about Gemini 2.5 Flash. I want it to be geared towards a technical audience." The first time that you mention the chatbot, if it's within a channel that you haven't interacted with the chatbot before, you'll have this prompt where researcher 2.0 know is not in the channel and then we can just go ahead and add it. Now that the assistant is added to our Slack channel here, the way that it will work is every time that you invoke a request to one of the chatbots that you set up, you you will get this reaction of the eyes and then you can take a look at the thread. So within here, I can see it's going through and it's scraping a number of different tools. It took the steps to research all of the information that I needed to know about Gemini 2.5 Flash. It scraped various different web pages based on that MCP server that we have for fire crawl. Within here, I can see the very detailed response that it's gathering towards us. As you can imagine, RunBear really shines in team settings because, as you can imagine, if you're trying to research a particular topic or whatever it might be, whatever the assistant that you've built the use case for. By being able to have this and reference it within a tool that you're already using like Slack is very convenient cuz within here, you can have conversations around different ideas. You can also invoke multiple assistants within here. And within here I can see there's even some Python examples, JavaScript examples. Where this can be particularly helpful for me is both researching particular topics with different individuals is I'll be able to bounce different ideas and be able to share different aspects about potential concepts that I have for different videos. Once you have your chatbot set up, you can always go back to the interface within Run Bear. There is a playground where you can try out what you've set up without having to go in the specific integration for whether it's Slack or Discord or Microsoft Teams. You do also have the ability to control where these different assistants are available. For instance, if you want it available to all of the Slack channels, you can set it to public. Alternatively, you can set it up in a way where it will only be within a specific channel if you'd like. One thing specifically that I want to call out is the smart reply feature. What this will allow you to do is to automatically send a private response when a relevant answer is found. Just to demonstrate how this would work is when you ask a particular query and you have smart reply enabled. What the assistant will do is say if there's relevant context for a question that you might just be asking the broader channel, it will give you a response that's only going to be visual to you unless you actually share it to the thread. For instance, if you ask a question like, "Does anyone know the contact person at Acme?" We can see the suggested response based on the conversation thread knowledge base. However you've set up the assistant, we can see the contact person at Acme is Kristen Brown. Similar to something that Sam Alman said in a recent blog post where he said, "We believe that in 2025 we may see the first AI agents quote unquote join the workforce and materially change the output of companies." Smart reply is a really clear example of exactly that. Smart reply, it enables you to have an AI agent that feels like a teammate. It's ready to jump in with helpful suggestions without needing to be explicitly mentioned. Now, in terms of use cases, there's a really rich set of examples on the Run Bear site. Everything from being able to query and analyze Air Table data all the way through to triggering actions within Zapier. You can also filter through a number of the different use cases that they have within here. So, in terms of really being able to get the most out of your AI teammate, now while it's going to be unique to whatever you set up, I encourage you to check out these use cases cuz there are a ton of really great examples that could potentially give you ideas in terms of how you can alleviate some of the work for both yourself as well as your team. I'll link everything within the description of the video. If you found this video useful, please comment, share, and subscribe. Otherwise, until the next