I agree, Jenny! 💯 It feels like we are just scratching the surface of what happens when the AI can actually reach out and touch the rest of our tech stack.
And at the same time, I'm curious to see which MCP servers end up being the "must-haves" and which ones are just noise. 🤔
Are you already building any, or just exploring the ecosystem for now? :)
Thanks for sharing that guide, Jenny! I love that you are going the Next.js route. It makes so much sense for getting these servers live without over-complicating the stack.
I will definitely take a look at how you are structuring yours. I am curious though, are you finding that Next.js makes it easier to handle the "live" data part of the MCP, or are you mostly using it for the ease of deployment? :)
Hey this is great piece on building MCP for context. However I was thinking I am doing the same context builing using Project for Claude Chat and the folders in my local for Claude code. How is MCP different?
Great question Meenal! Claude Projects are like a "backpack" where you pack files for one chat. MCP is more like a "live bridge" that lets the AI reach out and grab your logic dynamically whenever it's needed.
The advantage is that it transforms your static files into a live system that can communicate with your other tools. While Projects use static snapshots, MCP uses live data. MCP is for "doing" because an MCP server can allow the model to actually edit files or trigger API calls.
There's also a key difference in token usage. Projects are limited by the context window size (200k tokens), but MCP can theoretically access terabytes of data because it only fetches what is relevant to the prompt. I don't see myself using that much yet, but the potential is there!
Love how you did it Elena! I strongly believe that the future will be inundated (in a good way) by MCPs 😁
I agree, Jenny! 💯 It feels like we are just scratching the surface of what happens when the AI can actually reach out and touch the rest of our tech stack.
And at the same time, I'm curious to see which MCP servers end up being the "must-haves" and which ones are just noise. 🤔
Are you already building any, or just exploring the ecosystem for now? :)
I’m also building the MCP, would love to try out yours!
There are a few options, but my favorite ones are built via nextjs.
Just realized you can have access to it:) https://buildtolaunch.substack.com/p/how-to-build-mcp-server-complete-guide?utm_source=publication-search
Thanks for sharing that guide, Jenny! I love that you are going the Next.js route. It makes so much sense for getting these servers live without over-complicating the stack.
I will definitely take a look at how you are structuring yours. I am curious though, are you finding that Next.js makes it easier to handle the "live" data part of the MCP, or are you mostly using it for the ease of deployment? :)
You know me, now the most important part is the ease of deployment :)
Nice! Ok now I need to do it :)))
Thanks, Ileana! I'd love to see how you do it. :)
Hey this is great piece on building MCP for context. However I was thinking I am doing the same context builing using Project for Claude Chat and the folders in my local for Claude code. How is MCP different?
Great question Meenal! Claude Projects are like a "backpack" where you pack files for one chat. MCP is more like a "live bridge" that lets the AI reach out and grab your logic dynamically whenever it's needed.
The advantage is that it transforms your static files into a live system that can communicate with your other tools. While Projects use static snapshots, MCP uses live data. MCP is for "doing" because an MCP server can allow the model to actually edit files or trigger API calls.
There's also a key difference in token usage. Projects are limited by the context window size (200k tokens), but MCP can theoretically access terabytes of data because it only fetches what is relevant to the prompt. I don't see myself using that much yet, but the potential is there!
Let me know if that helps to clarify things :)