The Hands-On Lab
Stop reading. Start touching. Try MCP in 15 minutes.
Why This Lab Exists
You have read three modules about MCP strategy. You understand it conceptually. But you have never actually seen it work. This module fixes that. No code required.
Product people make better decisions about technology they have actually used. You would not spec a mobile app without using a smartphone. You would not commission an API without seeing one work. Do not spec an MCP integration without experiencing one.
This lab is designed to take 15 minutes. By the end, you will have touched MCP with your own hands, seen it in action, and understood the user experience of an AI assistant with MCP capabilities.
Lab 1: MCP in 60 Seconds (Claude Desktop)
The fastest path from "never touched MCP" to "just used it."
1 Open Claude Desktop
If you do not have Claude Desktop, download it from claude.ai/download. It is free.
2 Go to Settings, then Extensions
In the left sidebar of Claude Desktop, click Settings (gear icon). Look for "Extensions" or "Connected Tools."
3 Click "Browse extensions"
Claude Desktop will show you available MCP servers you can install. These are curated by Anthropic.
4 Install any extension
Pick one that looks useful to you. Examples: a file system tool (browse your computer), a web search tool, or a productivity integration. Click install.
5 Go back to chat and use it
Start a new chat. Ask Claude to do something using the tool you just installed. Examples:
- If you installed file system tool: "List the files on my desktop"
- If you installed web search: "Search for the latest news on MCP"
- If you installed a productivity tool: "Create a reminder for tomorrow"
Watch as Claude uses the tool you just gave it. That tool is an MCP server. That conversation is how AI integration works in 2026.
You just expanded what an AI can do by installing a plugin. That plugin is an MCP server. That is the entire concept. Everything else in this course is details about how to build and manage them.
Lab 2: Browse the Ecosystem (5 Minutes)
Understand the scale and variety of what has already been built.
1 Visit Smithery.ai
Go to smithery.ai. This is the largest MCP server registry.
2 Browse by category
Explore servers for different categories. Look for:
- Slack integrations
- GitHub integrations
- Salesforce integrations
- HubSpot integrations
- Google Drive integrations
- Database tools
3 Pick any server, read the description
Click into a server that interests you. You will see: a description of what it does, the Tools it exposes, the Resources it provides, and how many times it has been installed.
4 Visit mcp.so
Go to mcp.so. This is another registry. Search for your product category (e.g., "project management," "CRM," "analytics").
5 See who in your space already has an MCP server
The companies you see here are the ones already accessible to AI assistants. The ones you do not see are invisible.
This is the ecosystem your product either participates in or gets left out of. There are already 17,000+ MCP servers indexed. Every one of these servers represents a product that AI assistants can use. If yours is not on the list, you are invisible to a growing segment of potential users.
Lab 3: The Playground (5 Minutes)
Get your hands on MCP Tools in a browser-based sandbox.
1 Visit mcpshowcase.com
This is a web-based MCP playground. No installation required.
2 Connect to a demo MCP server
The playground will prompt you to select a server to connect to. Pick one from the list.
3 Chat with Claude through the playground
Ask Claude to use the MCP Tools available. Watch as Claude calls the Tool, gets a response, and reports back to you. Watch as different MCP servers give Claude different capabilities.
4 Try different servers
If you have time, connect to multiple servers. See how the same AI gains different capabilities depending on which MCP servers are connected.
An AI's capabilities are modular. Connect a different MCP server, get a different capability. This is what "MCP Surface Area" means in practice. The more Tools and Resources you expose, the more useful AI assistants become when working with your product.
Lab 4: MCP Apps (5 Minutes)
MCP is not just text responses. It can render rich interactive interfaces.
1 In Claude Desktop, look for "MCP Apps" in settings
The MCP Apps extension enables Tools to return full interactive interfaces (dashboards, forms, charts) directly in conversation.
2 Ask Claude to use a Tool that returns visual data
If you have an MCP server installed that provides charts, dashboards, or forms, ask Claude to use it. Examples:
- "Show me my sales forecast as a chart"
- "Generate a proposal form I can fill in"
- "Create an interactive budget dashboard"
3 Watch the interface render in the chat
You will see full interactive components appear directly in the conversation. This is more powerful than traditional API integrations.
MCP integrations are not limited to "read data, write data." They can power entire interactive experiences inside AI conversations. Think: a customer support agent that pulls up a live dashboard, or a sales assistant that generates a populated proposal form. The UX possibilities are much larger than most teams realise.
Lab Debrief: Three Questions for Your Team
After completing the labs, write down your answers to these three questions. Bring them to your next team meeting or sprint planning session.
Think back to the Smithery or mcp.so servers you browsed. Which of your product's features would be most useful if an AI assistant had access to them?
Write down 3-5 features that would unlock value for your users if they could delegate them to an AI.
During Lab 2, you searched your category. What did you find? Which companies are already visible to AI assistants?
Write down the companies you found. This is your competitive landscape for the AI layer.
Start small. Do not think about your entire product. Think about the narrowest possible MCP server that would still be useful.
Write down one resource you could expose (read-only data from your product) in the next two weeks. This becomes your MVP.
Next Steps After the Lab
If you answered the three debrief questions above, you have enough information to start a conversation with your engineering team about building an MCP integration. You do not need to understand the code. You understand the what and why. Your team will figure out the how.
Bring your debrief answers to your next product meeting. Share what you learned from the lab. Show your team some servers from Smithery or mcp.so that are similar to your product. Ask: "Could we build something like this?"
That conversation is often the start of an actual MCP roadmap.
After trying MCP hands-on, what changed about how you think about your product's AI integration strategy?
Key Takeaways
- Hands-on experience beats theoretical knowledge. Fifteen minutes of actually using MCP tools teaches you more than hours of reading about them.
- The ecosystem is already here. Smithery.ai and mcp.so show you the scale of what has been built. Seeing it makes the competitive pressure real.
- MCP is modular. Different servers give AI different capabilities. Your job is to expose the capabilities that matter most to your users.
- The Surface Area Mapper is now concrete. After playing with different MCP servers, you can envision what your own server would expose.
- You are ready to start planning. The three debrief questions are enough to brief your engineering team and start a real MCP roadmap.