Lab Guide — MCP Concepts & Protocol

Connect to an MCP server and discover its capabilities

Using the MCP Python SDK, create a client session that connects to a server process and lists its available tools.

1. Create an MCP ClientSession connected via stdio transport
2. Call initialize() to perform the capability handshake
3. Call list_tools() and print each tool’s name, description, and input schema

# Expected output format:
# Tool: read_file
#   Description: Read the contents of a file
#   Parameters: {"path": {"type": "string"}}
# Tool: list_directory
#   Description: List files in a directory
#   Parameters: {"path": {"type": "string"}}

Checkpoint: You can list all tools from the connected server with their schemas.

Execute a tool call through the MCP protocol

Use the client session to call a tool and process the result.

1. Call a tool using session.call_tool(name, arguments)
2. Parse the response content
3. Handle errors (invalid tool name, missing parameters)

# Call a tool
result = await session.call_tool("read_file", {"path": "/tmp/test.txt"})
print(result.content[0].text)

Checkpoint: You can call a tool, receive its result, and handle error cases.

Research 5 popular MCP servers

Survey the MCP ecosystem and document 5 servers. For each, record:

• Server name and source (official vs community)
• What tools it exposes
• What resources it provides (if any)
• Transport type (stdio / SSE)
• A practical use case for the server

Compile your findings into a comparison table in the notebook.

Checkpoint: A completed table with 5 servers, their capabilities, and use cases.