MCP Course Home → Module 4 of 7

You have spent three modules reading about MCP strategy. You understand the business case, the decision framework, and the building blocks. But you have never actually seen MCP in action.

Product people make better decisions about technology they have used. You would not spec a mobile app without owning a smartphone. You would not design a checkout flow without buying something online. Do not make MCP decisions without experiencing MCP.

This module is 15 minutes of hands-on work. No code. No terminal. No engineering degree required. By the end, you will have installed an MCP extension, browsed the ecosystem, tested tools in a playground, and seen interactive MCP in action.

What you will do: Install an MCP extension in Claude Desktop and use it.

Steps:

What just happened: You expanded an AI assistant's capabilities by installing a plugin. That plugin is an MCP server. It exposed new tools to Claude, and Claude used them on your behalf.

This is the entire MCP concept in action. A product (in this case, a web search service or file system) made its functionality available through MCP. An AI assistant (Claude) discovered that functionality and used it. No custom integration. No API key exchange. One click.

Everything else in this course is about helping you do this for your own product.

What you will do: Understand the scale and breadth of what has already been built.

Steps:

What to notice: Every server in these directories represents a product that AI assistants can interact with. Every one of them is discoverable. When a user asks an AI assistant to "do something with [category]," these are the products that come up.

Write down two things: (a) whether any of your competitors already have an MCP server, and (b) how many servers exist in your product category. You will need these answers in Module 5 when you build the business case.

If you have Claude Desktop, these MCP extensions demonstrate the breadth of what MCP can do. Install 2 or 3 that match tools you already use.

Productivity and Communication: Slack MCP — Automate team communication, summarise threads, draft replies, search channels. If your team lives in Slack, this shows how MCP bridges AI and daily workflows. Notion MCP — Access and manage your Notion workspace. Query databases, create pages, update notes. A strong example of AI-powered knowledge management. Gmail MCP — Email search and management through Claude's Connectors Directory.

Development and Version Control: GitHub MCP — The most widely used MCP server. Full repository management: commits, branches, pull requests, issues. Essential for understanding how MCP handles complex, multi-action integrations. Playwright MCP — Web automation and testing. Watch Claude interact with web applications through structured accessibility trees. Shows MCP's power for browser-based workflows.

Data and Databases: PostgreSQL MCP — Query databases naturally, manage schemas, analyse data. If your product has a database, this is what your MCP server will feel like. Supabase MCP — 20+ tools for table design, migrations, SQL queries. A rich example of a comprehensive MCP surface area.

Design and Creative: Figma MCP — Extract design specs, sync tokens, convert components to code. Demonstrates MCP bridging design and development workflows.

Search and Information: Brave Search MCP — Live web search with privacy focus, pagination, and freshness controls. Shows how MCP handles real-time external data.

System and Files: Filesystem MCP — Secure local file operations. Read, write, search, manage directories. The most fundamental MCP capability, and the easiest to understand.

What you will do: Use MCP tools in a browser-based sandbox.

Steps:

What to notice: The AI's capabilities are modular. This is the core MCP insight. The same AI becomes a different product depending on which MCP servers are connected. Your product can be one of those modules.

This is what we meant by "MCP Surface Area" in Module 3. The tools and resources you expose through MCP become capabilities that AI assistants can use. The more thoughtfully you design your surface area, the more useful AI becomes when it works with your product.

What you will do: See MCP return rich, interactive interfaces (not just text).

Steps:

What to notice: MCP is not limited to text-in, text-out. The MCP Apps extension enables tools to return full interactive interfaces: dashboards, forms, charts, and workflows rendered directly in the AI conversation. This means your MCP integration can deliver rich experiences, not just data retrieval.

Think about what this means for your product. A sales tool could render a pipeline dashboard inside a Claude conversation. A project management tool could display a Kanban board. A finance tool could show an interactive P&L statement. The AI conversation becomes a distribution surface for your product's interface.

You have now installed, browsed, played, and seen. Before you close this module, write down your answers to three questions. These go directly into your next sprint planning conversation.

1. "Which of our product's features would be most valuable as MCP tools?"

Think about what you saw in the playground. What made you think "we should have that"? What would your users ask an AI assistant to do with your product?

2. "Which competitor or adjacent product already has an MCP server?"

You checked mcp.so. What did you find? If a competitor is already there, you know the urgency. If nobody in your space is there, you know the opportunity.

3. "What is the first MCP integration we could ship in 2 weeks?"

Start small. One read-only resource that exposes useful data. Not the full surface area map. Just one thing. What would it be?

← Module 3: Anatomy of an MCP Integration | Course Home | Next Module: Selling MCP Internally →