Imagine you can turn call from an AI chat bot into super powerful AI agents that actually does things for you. This is exactly what is happening now with MCP model context protocol. So in this video I'll explain what is MCP, why this is such a big deal. Share my framework that you can apply immediately to turn Claude into an AI agent together with five exciting AI agents ideas you can start using today. Let's go. so don't let the term MCP scare you. In simple terms, it's just a technology treat it like a universal plug that offers a standard way for any AI models, including Claude, to connect to the outside data and tools. Without MCP, the only way Claude can access data was you uploading the file in the chat. Now with MCP Claude can directly access to any external data source and system by connecting to an MCP server. And this is exactly what makes Claude and MCP such a powerful combo because Claude can now retrieve so much more context and access to tools to perform the actual task for you. Now, I won't go through the details on how to connect to an MCP server. There already so many tutorials out there. I'll put all the resources below , it's not difficult to follow. What I actually want to highlight is generally there are four ways you can use MCP with Claude. First, the official native integration offer on Claude, like Google Drive calendar, Gmail. Just click connected apps and start the connection. Official MCP servers that offer the official integrations in case the native integrations is not available, which I always recommend because these are well maintained like Perplexity MCP, Notion MCP, they have their official MCP server. And then community built. MCP Servers. You may want to use this if the previous two options are not available like Figma MCP is a community built MCP server. And finally, self-built MCP server. If there is nothing you can find and you can build one yourself using no-code platforms like n8n, like Google Analytics, Google Doc, which is powered by my own self-built MCP server on n8n. Note for now, I recommend you to use Claude Desktop to have the best MCP support experience other than the native integrations, but I believe Anthropic will soon adding more MCP options to the web app as well. Before I move on, if after watching this video you find yourself wanting to get started with creating AI agents, I also recommend you check out these free playbook master AI agents from HubSpot. I've put the down link in the description. This is a practical guide to help you get started with AI agents. It covers everything from what are AI agents, how they work, the latest AI agent landscape. High impact use cases to a step-by-step implementation roadmap and common mistake you should avoid. You'll even get a practical checklist of how to use AI agents, whether you're just getting started or looking to scale up what you have already implemented. What I like is how it breaks down the implementation process in the different stages with clear criteria and examples that makes this feel much more doable, especially if you have no idea where to start. I think it makes it easy to follow along. Check it out in the description below is free to download. And thank you HubSpot for sponsoring this video. Alright, here is the framework to turn Claude into an AI agent. First, identify your problem and needs first. Is it research, managing emails or anything else? Do not start with the tools. Then determine what necessary tools to solve this problem and connect Claude with the right tools through the use of MCP or the build-in native integrations. For example, a perplexity MCP that enable Claude to do web search. And finally, using Claude project features, define the role and responsibilities, available tools, workflow details for this agent through the project instruction. So every Claude project is just like a specific AI agents that does certain tasks, and here are five amazing AI agents ideas that I built using this framework and you will know how to apply the same concept to build whatever AI agents that you need on Claude. Our first agent is a research agents. Research is what every professionals, marketers would need, and this is what AI can really help you to speed up. So we have already had different tools turned on, including perplexity, but to let you see the difference, I will make sure the native web search is turned off. So Claude can only use perplexity for search. And here is our research agents, and literally it's just a Claude project. Now, the most important part is the project instruction. So we will define the work scope, the tools that are available, the workflow details. So every time Claude will develop a research plan with three question similar to the experience on Chatgpt deep research, and then do the actual searching and then create a visual summary and also generate a report to Google Doc and return you with the link and also the command to trigger this workflow and any important guardrails for this agent. So let's try a simple research topic first. Let's ask the latest trend about minimalism. And if you want, you can just type the topic. Claude will be smart enough to pick this up. So through the process Claude will prompt you as it needs access to perplexity. You can click, allow always or allow once. Now you can see it will follow the instructions and generate two sets of beautiful interactive visual summary, recapping the key insights gathered from the research, which is something that I really like Claude about. Again, if you want, you can publish this artifact and share with your team. Now what amazing is that Claude will also return a Google documentation link, which I can access, and it is the report about the research topic with simple formatting and the list of source URLs. So this time let's try a more complicated example. This time I will also upload a company information. So in this case it is about Shopify. Upload to Claude and ask it to research a more controversial research topic, the impact of AI on leadership. So this time also turn on extended thinking because I also wanted to propose recommendations from the research. Again, you can see it will generate the visual summary, it’s a priority matrix, highlighting the areas Shopify should focus first, which is really useful, and this is really Claude strength for generating this kind of strategic insights. And also the report and this time is much more detailed and completed than the last one, although it is in a markdown format, but it's not a big deal. We can always enhance it through the prompts, And I really like how Claude breaks down the recommendations and also with an implementation roadmap using its thinking ability with a list of citations of the full URLs so this is absolutely amazing. Of course, this is simplified demo, but you can further refine instructions to this agent. Perhaps also prepare an infographic, do a more in-depth research plan, send you an email so there are so much more possibilities and potential for this research agent. The next agent is what I call a business intelligence agent. Basically, this is an agent that interpret all sorts of data around your business and marketing and to interpret insights. This is one of my favorite AI agents ideas because we always need to analyze so much data and extract insights. These agents will make your work so much easier. So I have set up a new project for this business intelligence agent. Again, for the project instructions. We define the role for this agent, the tools, which is Google Analytics and Gmail in this case, as most marketers, I know we can't live without web analytics, but if you have connected to other data sources, also specify here for Claude to use. Then the workflow details. Note for this agent. It can also send email for the GA report. Now let's ask it to analyze the session performance for the past 90 days, break down by channels, and then send an email to me. Then again, you can see it will trigger the connections. Click allow, and it will now pull the Google Analytics data and generate the visualization. Then it will also trigger the email sending and everything is automated. And you can see the dashboard with the callout boxes, is also following my instructions to use a purple color scheme and try to have multiple visualization. And this is the power of Claude for visual storytelling. And Claude has also sent a report to my mailbox automatically, with the rich formatting, callout boxes, key insights and recommendations. So this is absolutely amazing, Imagine how much time it can save you. You can just ask whatever metrics that you want to check from a data source and then Claude will summarize the insights for you and generate reports and even send you the email quick summary. So definitely try this agent ideas yourself And you can even build an automation by using a command for this agent. So every time it will know exactly what to do and auto generate a report for you. So think about what data you might need to interact all the time. Search trend data, keyword data, third party data, and ask Claude to do the storytelling for you and this is a game changer. The next is a personal workflow assistant. So this agents will connect to your Gmail, Google Calendar, or other productivity apps to streamline your daily work process, retrieve context from your files, just like your personal assistant, but note I will be using my own-built MCP server on n8n for the Gmail integrations, so Claude can also able to send emails and not just read emails. *** Again, I already have this agent set up. So here are the instructions for this agent, and it's very important to define what tools this agent should use, like Gmail and Calendar in this case, and what the workflow process looks like, when they should generate visual summary or to draft you communications. So let's say I want to retrieve everything about a mentorship program I have done last year with my mentee, Jessica. And I wanted it to generate a simple project dashboard to recap the key project milestone and status. So you will see it will trigger the connections and calling the MCP server I set up on n8n. So click allow. Okay, now it is searching the emails and also the calendar related to this project and retrieving all the context for you. So immediately you will have the project summaries right in the chat with a beautiful dashboard. The program overviews, the project timelines, key milestone about the project, key actions made. So this is super useful whenever you just want to retrieve context about certain project. Especially it has started long time ago that you want to immediately get a summary of what is going on. and this is the key difference compared with Gemini. Although Gemini has the best integrations right within Gmail, but it can't generate any visualization. Same case using Gemini Advanced. I can even ask it to send this as a quick email summary for me, everything is automated that’s really powerful use of this Claude personal assistant. You can also ask Claude to check anything for you. For example, I can ask it to help me to check any leads I haven't follow up in the past seven days. And it would immediately summarize the status for me. Who are they? And even with little context for me to make a quick decision, and this is something I found even Gemini can't do well for now. And it can also summarize like newsletter content for me if I'm just too busy to read them all. And this is really Claude strength, very detail oriented, and you can just use Claude as your personal assistant. And these are just some of the use cases. You can also ask it to audit your calendar, mailbox, visualize how you spend your time, provide you suggestions, and not just simply booking meetings because this is something Gemini can already do for you. But what makes Claude different is how it can always visualize the meaningful insights for you. ****** The next UX design agent, I must say that UX design is an increasingly important part of every growth marketing strategy. So turn Claude into a simple UX design agent will make this very useful, so you can streamline lot of design analysis process, prototyping by connecting it to Figma. So here is our UX design agent. So in the project instructions, we specify tools like Figma MCP, to access the design file on Figma. Google Drive to access the user feedback, and then the workflow process details for each type of tasks, like code generation, prototype creation, design analysis and other important guardrails for this agent, like a Figma link should always be included. So let's say I have this eStore product page mockup design, and I want Claude to give me improvement suggestions. So first, copy this Figma design file link and paste to the chat. We are not done yet. Let's say here is the real user feedback from our research process for this Sunshine eStore project. And we also want Claude to consider these user feedback when generating the recommendations. So now use this prompt to ask it to analyze all the materials, including the Google Doc on our Google Drive and suggest prioritize action. I also suggest, turn on thinking mode for this task so you can get more thoughtful recommendations with its thinking ability. So you can see it will based on the mockup and feedback to come up with high priority improvement. And it even visualize a rough mockups for its suggestions. And this is why I always love Claude about, and you can see it is addressing the feedback from users in this mockup. Now let's say this is our official e-store style guide on Figma. Then we can just ask Claude to create a new product page prototype and following the brand style guideline. So this is absolutely amazing. It's on brand tone. With the improvement made, you can just download the SVG file and we import to Figma and then you can share with your team to immediately visualize the side by side improvement. So this will save you so much time. *The next project intelligence agents, or you can call it knowledge agent. I know most marketers love using Notion or Airtable as their team shared knowledge repository or keep track of different project datas. So by connecting with Notion, we can turn Claude into a powerful project agent that help you to analyze the project data and to actively update these notion projects. So this is our project intelligence agent. For this agent. We will connect it with Notion and perplexity. So we can also enable this agent to do some research tasks. Again, include the workflow details, like how it should follow the notion template when generating a new page. Now let's say we have this SWOT analysis project on Notion It is used it to store all kinds of SWOT analysis we do for our competitors. And this is the template. So first, use this prompt to ask Claude to perform a simple SWOT analysis about the brand Canva. So you can see it will trigger the connections with perplexity for the web search and notion to automatically generate the new SWOT analysis on this notion project I just show you . And you can see it will also follow instructions to generate the artifact for this task, which you can later publish or export for your teammate. And the best thing is Claude will automatically insert a new SWOT analysis to this Notion Project where you can click on it and it is following the same template. So imagine how much easier it will make your work. So you no longer need to copy and paste or export and import. So this agent will handle all that for you. And now another example. Let's say we have this project planner on notion. It's just a demo. It has all the project and task status details. So we can use this prompt to ask Claude to analyze this project planner and visualize the project status. As well as completion rates and resource allocation. So unlike retrieving your unstructured data on your personal mailbox, these are structured data and the visualization is to be shared across your team. So how the project status is going, resource workload, task priority, and most importantly, the potential bottlenecks and the dependency. So it will give you immediate insights about a ny active projects that you're collaborating with your internal and external team. If you don't want visuals, you can just ask Claude any question about any projects and it will immediately pull insights for you. And same time, you can also ask it to update the project. For example, I can ask it to add a new task on this project planner. And note, I didn't use any structured format, is completely conversational and Claude is smart enough to pick up for you and understand that and populate the task details on this planner. So this is a huge time saver. But keep in mind, the more the autonomy of these AI agents, the higher are the potential risks. So for high stake actions, always have an approval mechanism or have humans in the loop before these AI agents take the actual actions. Claude and MCP is a real game changer that you shouldn't miss. And the thing is setting up MCP on Claude can be very approachable even if you have little or no technical background. The effort is worth the spend if you want to streamline your workflow further. If you want more inspiration, you can join my community. You'll get access to my other feature content and all my tested prompt I share on my channel. You can find the link in the description to join, and before you go, I highly recommend you to watch this video about AI agents so you have a better concept when you should use AI agents. I will see you next time.