Skip to content
AI Systems 2026

Lab 03: MCP Server Implementation

Elementary Due: 2026-03-24

Objectives

Section titled “Objectives”
  • Implement an MCP server with FastMCP (Tool + Resource + Prompt — all three primitives)
  • Validate the server with MCP Inspector
  • Apply input validation and path traversal defense
  • Configure filesystem + Git MCP servers

MCP Overview

Section titled “MCP Overview”

MCP is a protocol that allows AI coding agents to communicate with external tools and data sources in a standardized way. Capabilities are exposed through three primitives — Tool (invoked by the model), Resource (controlled by the application), and Prompt (selected by the user).

AI Coding Agent
↓
MCP Client
↓
MCP Server (Tool + Resource + Prompt)
↓
Filesystem / Git / Custom API

Implementation Requirements

Section titled “Implementation Requirements”

1. MCP Server Configuration — settings.json

Section titled “1. MCP Server Configuration — settings.json”

Add built-in servers to your AI coding CLI’s configuration file.

Config file: ~/.claude/settings.json (or project-level .claude/settings.json)

 
{
  "mcpServers": {
    "filesystem": {
      "command": "npx",
      "args": [
        "-y",
        "@modelcontextprotocol/server-filesystem",
        "/home/[student-id]/lab-03"
      ]
    },
    "git": {
      "command": "npx",
      "args": [
        "-y",
        "@modelcontextprotocol/server-git",
        "--repository",
        "/home/[student-id]/lab-03"
      ]
    }
  }
}

2. Environment Setup and Server Verification

Section titled “2. Environment Setup and Server Verification”

  1. Project Structure and Environment Setup

    Terminal window
    mkdir -p lab-03-mcp && cd lab-03-mcp
    python -m venv .venv
    source .venv/bin/activate
    

    # Using uv (recommended)
    uv add "mcp[cli]" fastmcp
    

    # Using pip
    pip install fastmcp
    lab-03-mcp/
    ├── custom_server.py       # FastMCP server (Tool + Resource + Prompt)
    ├── settings.json           # MCP configuration
    └── test_sample.py          # Tests to run via the custom server's run_pytest

  2. Check Built-in Server List

    Terminal window
    claude mcp list
    # filesystem  npx -y @modelcontextprotocol/server-filesystem ...
    # git         npx -y @modelcontextprotocol/server-git ...

  3. Test the Filesystem Server

    Terminal window
    claude "Show me the list of files in the current directory."
    # The list_directory tool from the filesystem server should be called

  4. Test the Git Server

    Terminal window
    git init /home/[student-id]/lab-03
    claude "Summarize the last 3 commit messages."

3. Custom MCP Server — custom_server.py

Section titled “3. Custom MCP Server — custom_server.py”

Implement a server using FastMCP that includes all three primitives (Tool + Resource + Prompt).

custom_server.py
import os
import subprocess
import sys
from fastmcp import FastMCP


mcp = FastMCP("lab03-custom", description="Lab 03 Practice Server")


# === Allowed path configuration (path traversal defense) ===
ALLOWED_BASE = os.path.realpath(os.path.dirname(__file__))




def _validate_path(path: str) -> str:
    """Defends against path traversal attacks. Blocks access outside the allowed directory."""
    real_path = os.path.realpath(path)
    if not real_path.startswith(ALLOWED_BASE):
        raise ValueError(
            f"Access denied: {path} — outside allowed scope ({ALLOWED_BASE})"
        )
    return real_path




# === Tool 1: Run pytest ===
@mcp.tool()
def run_pytest(path: str) -> str:
    """Runs pytest in the specified directory and returns the results.


    Args:
        path: Directory path where pytest should be run
    """
    safe_path = _validate_path(path)
    try:
        result = subprocess.run(
            ["python", "-m", "pytest", safe_path, "-v", "--tb=short"],
            capture_output=True,
            text=True,
            timeout=60,
        )
        return result.stdout + result.stderr
    except subprocess.TimeoutExpired:
        return "ERROR: pytest execution timed out (60 seconds)"
    except Exception as e:
        print(f"[ERROR] {e}", file=sys.stderr)
        return f"ERROR: {e}"




# === Tool 2: Count lines ===
@mcp.tool()
def count_lines(file_path: str) -> str:
    """Returns the number of lines in a file.


    Args:
        file_path: Path to the file to count lines in
    """
    safe_path = _validate_path(file_path)
    try:
        with open(safe_path, "r") as f:
            lines = f.readlines()
        return f"{len(lines)} lines"
    except FileNotFoundError:
        return f"ERROR: File not found — {file_path}"
    except Exception as e:
        print(f"[ERROR] {e}", file=sys.stderr)
        return f"ERROR: {e}"




# === Resource: Project statistics ===
@mcp.resource("project://stats")
def project_stats() -> str:
    """Returns file statistics for the project directory."""
    py_files = []
    for root, _dirs, files in os.walk(ALLOWED_BASE):
        for f in files:
            if f.endswith(".py"):
                py_files.append(os.path.join(root, f))
    total_lines = 0
    for fp in py_files:
        try:
            with open(fp) as f:
                total_lines += len(f.readlines())
        except Exception:
            pass
    return (
        f"Python files: {len(py_files)}\n"
        f"Total lines: {total_lines}\n"
        f"Project path: {ALLOWED_BASE}"
    )




# === Prompt: Code review ===
@mcp.prompt()
def code_review(file_path: str) -> str:
    """A structured prompt template for requesting a code review.


    Args:
        file_path: Path to the file to review
    """
    return (
        f"Please code review the following file: {file_path}\n\n"
        f"Review criteria:\n"
        f"1. Security vulnerabilities (path traversal, command injection, input validation)\n"
        f"2. Error handling (exception handling, safe failures)\n"
        f"3. Code quality (readability, naming, structure)\n"
        f"4. Testability (is the design easy to test?)"
    )




if __name__ == "__main__":
    mcp.run()

4. Register the Custom Server

Section titled “4. Register the Custom Server”

Add the custom server to settings.json.

{
  "mcpServers": {
    "filesystem": { "...": "..." },
    "git": { "...": "..." },
    "custom": {
      "command": "python",
      "args": ["/home/[student-id]/lab-03-mcp/custom_server.py"]
    }
  }
}

5. Validate with MCP Inspector

Section titled “5. Validate with MCP Inspector”
Terminal window
# Run Inspector
npx @modelcontextprotocol/inspector python custom_server.py


# Go to http://localhost:6274 in your browser, then:
# 1. tools/list → verify run_pytest, count_lines
# 2. resources/list → verify project://stats
# 3. prompts/list → verify code_review
# 4. tools/call → run run_pytest and check the result
# 5. Capture screenshots (for assignment submission)

6. Verify Custom Tool Behavior

Section titled “6. Verify Custom Tool Behavior”
Terminal window
claude "Run pytest in the lab-03-mcp directory."
# The run_pytest tool should be called

Deliverables

Section titled “Deliverables”

Submit a PR to assignments/lab-03/[student-id]/:

  • settings.json — Configuration file with filesystem, git, and custom servers registered
  • custom_server.py — FastMCP-based implementation with all three primitives (Tool + Resource + Prompt)
  • test_sample.py — Test file to be run via the custom server’s run_pytest (minimum 3 tests)
  • Inspector screenshots — Confirming tools/list, resources/list, prompts/list
  • README.md — Verification results for each MCP server, security validation (path traversal blocking test), troubleshooting notes