This week I had the epiphany that AI in VS Code is about WAY more than just writing code.
MCP Servers rethink the entire dev workflow...
The agent can use playwright MCP directly - no tests required - to open a browser and diagnose an issue in your app. You can then turn right around and have the GitHub MCP server create the issue for you, complete with all the context and reproduction steps.
I was not prepared for how smooth this was. It was one of those moments.
Curious — what are you using MCP for beyond just writing code?
MCP servers are really powerful for improving your workflows in ways that I never even saw coming. So let me show you what I mean. In my site here for sharing lists of links, I'm pulling the open graph information from the website as I'm adding the link and you can see that some websites just don't work. So here's one for Battle.net. And if you know what that is, what's up fellow Overwatch players? But I want to show you how MCP can help with this. So let's go back to Visual Studio Code here. So here in Visual Studio Code, what I want to do. Here is just tell it that the OG information for Battle.net isn't showing up and use a Playwright MCP tools to figure this out. So you can see I've got Playwright installed here in my MCP. It's right here playwright. And So what it then done is it's now navigated to the URL and then got a page snapshot which gives it all the HTML so that it can look at it and explain what's happening. So it says I can see what's happening. It's actually we're actually being redirected. And so it looks like the redirect page is highly dynamic, but with modern web technologies. So apparently, apparently the page isn't rendering when we're trying to scrape it. It must be being rendered by JavaScript. And then it gives us some ways that we can handle that. Now at this point, I don't actually want to do this right now, but it's an issue I need to do in the future. So check this out. That's great. Thank you so much for the explanation. Can you go ahead and open an issue on GitHub for this? So in this case, I don't even have to open the issue, right? Like I could just give this back to the agent and be like, great, put that description in GitHub and I'll work on this at a different time. And this is incredible because it can actually write the issue for us, including all of the information that I'm going to need later on or that maybe copilots gonna need later on to actually solve this. And it does this using the GitHub MCP server. So it's actually created the issue here and then it gives us a link to it. So let's check it out here. It is root cause impact. Post solutions, technical details, I mean just all the information that you would need to solve this. MCP is incredibly powerful and it's completely changing the way that I work with AI. What are you doing with it? I'd love to know, Let me know in the comments.
Find me at Web Summit Vancouver | Principal Instructor at LinkedIn | Pragmatic AI Explainer | Tech Educator | TEDx Speaker | Neurodivergent System Thinker | Dad
MCP is both a huge step forward for automation and a huge step into a minefield of novel security and activity issues. We have a very short runway to figure out how to add some sort of governance layer with vetting and security certs to this protocol (maybe something similar to npm), otherwise the wild west I'm already seeing in this space will turn into a catastrophic mess.
I feel a bit like we're at a sharp bend in raging whitewater rapids, and while some of us up front are shouting about the very loud roar of a waterfall just ahead, the people in the back have thrown all the life vests overboard and are strapping rockets to the raft to make it go faster.
Lead Software Engineer | Azure Cloud & AI Solutions | Agile & DevOps Expertise | Tech for Social Impact
1wIt was great seeing this in your session today at Build!