MCP Tools Hands-On: 5 Open Source Powerhouses to Connect AI Agents to the World

1. Why Do You Need MCP Tools?

Before MCP protocol appeared, making AI access external data was like groping in the dark:

Each tool had its own API format
Authentication methods varied wildly (OAuth, API Key, JWT…)
Error handling had no unified standard
Debugging? Basically guesswork

MCP changed all that. It defines a standard interface, allowing any AI model to interact with external tools in the same way. Just like USB-C unified charging interfaces, MCP is unifying how AI connects to the world.

2. 5 Must-Try MCP Open Source Tools

1️⃣ MCP FileSystem Server - Let AI Read and Write Your Files

GitHub: https://github.com/modelcontextprotocol/servers/tree/main/src/filesystem

This is one of the most basic yet most practical MCP tools. Once installed, AI can safely access files in specified directories.

Installation Steps:

# Install with npm
npm install -g @modelcontextprotocol/server-filesystem

# Create config file ~/.mcp-config.json
{
  "mcpServers": {
    "filesystem": {
      "command": "npx",
      "args": ["-y", "@modelcontextprotocol/server-filesystem", "/home/yourname/documents"],
      "env": {}
    }
  }
}

Feature List:

read_file - Read file contents
write_file - Write to files
list_directory - List directory contents
search_files - Search files
create_directory - Create directories

Usage Example:

# AI can call through MCP
result = mcp_client.call_tool("filesystem", "read_file", {
    "path": "/home/yourname/documents/notes.md"
})
print(result.content)

Security Tip: Only expose necessary directories, never mount the root directory /!

2️⃣ MCP PostgreSQL Server - Natural Language Database Queries

GitHub: https://github.com/modelcontextprotocol/servers/tree/main/src/postgres

Let AI query databases directly using natural language, no need to write SQL.

Installation and Configuration:

npm install -g @modelcontextprotocol/server-postgres

# Config file
{
  "mcpServers": {
    "postgres": {
      "command": "npx",
      "args": ["-y", "@modelcontextprotocol/server-postgres", "postgresql://user:pass@localhost:5432/mydb"],
      "env": {}
    }
  }
}

Real-World Scenario:

User: Find the top 10 products with highest sales from last month

AI (via MCP):
  → Call postgres.read_query
  → Auto-generate and execute SQL
  → Return structured results

Supported Operations:

Execute read-only queries (safe mode)
Get table structure information
List all table names
Parameterized queries to prevent SQL injection

3️⃣ MCP GitHub Server - Intelligent Repository Management

GitHub: https://github.com/modelcontextprotocol/servers/tree/main/src/github

Let AI help manage your GitHub repositories, from viewing Issues to creating PRs.

Installation Steps:

npm install -g @modelcontextprotocol/server-github

# Need GitHub Personal Access Token
# Visit https://github.com/settings/tokens to create one
# Permissions: repo, read:user, user:email

{
  "mcpServers": {
    "github": {
      "command": "npx",
      "args": ["-y", "@modelcontextprotocol/server-github"],
      "env": {
        "GITHUB_PERSONAL_ACCESS_TOKEN": "ghp_xxxxxxxxxxxx"
      }
    }
  }
}

Core Features:

search_repositories - Search repositories
get_issue - Get issue details
create_issue - Create new issue
list_pull_requests - View PR list
get_file_contents - Read file contents
create_branch - Create new branch

Automation Example:

Scenario: Automatically organize Issues

AI Workflow:
1. Call github.list_issues(repo="myproject", state="open")
2. Analyze content and labels of each issue
3. Call github.update_issue() to add classification labels
4. Create summary document for high-priority issues

4️⃣ MCP Puppeteer Server - Web Scraping and Automation

GitHub: https://github.com/modelcontextprotocol/servers/tree/main/src/puppeteer

Enable AI to access real-time web content, perform data collection and automation.

Installation and Configuration:

npm install -g @modelcontextprotocol/server-puppeteer

{
  "mcpServers": {
    "puppeteer": {
      "command": "npx",
      "args": ["-y", "@modelcontextprotocol/server-puppeteer"],
      "env": {}
    }
  }
}

Key Features:

puppeteer_navigate - Open web page
puppeteer_screenshot - Take page screenshot
puppeteer_click - Click element
puppeteer_fill - Fill form fields
puppeteer_evaluate - Execute JavaScript

Real-World Case: Competitor Price Monitoring

# AI executes automatically
pages = [
    "https://example.com/product/1",
    "https://example.com/product/2"
]

for url in pages:
    mcp.call("puppeteer", "navigate", {"url": url})
    content = mcp.call("puppeteer", "evaluate", {
        "script": "document.querySelector('.price').textContent"
    })
    # Record price data

5️⃣ MCP Git Server - Version Control Automation

GitHub: https://github.com/modelcontextprotocol/servers/tree/main/src/git

Let AI understand and operate Git repositories, enabling intelligent code management.

Installation:

npm install -g @modelcontextprotocol/server-git

{
  "mcpServers": {
    "git": {
      "command": "npx",
      "args": ["-y", "@modelcontextprotocol/server-git"],
      "env": {}
    }
  }
}

Supported Commands:

git_status - View repository status
git_diff - View code changes
git_log - View commit history
git_commit - Create commits
git_branch - Manage branches

Intelligent Commit Example:

Scenario: Auto-generate Commit Messages

AI Workflow:
1. Call git.diff() to get changes
2. Analyze change types (feature/fix/refactor)
3. Generate Conventional Commits-compliant message
4. Call git.commit() to execute commit

3. Quick Start: Set Up Your First MCP Agent in 10 Minutes

Step 1: Install MCP Host

Recommend using Claude Desktop or custom Host:

# Use official CLI
npm install -g @modelcontextprotocol/cli

# Or use Python
pip install mcp

Step 2: Configure MCP Servers

Create ~/.config/claude/mcp.json:

{
  "mcpServers": {
    "filesystem": {
      "command": "npx",
      "args": ["-y", "@modelcontextprotocol/server-filesystem", "/home/yourname/projects"]
    },
    "git": {
      "command": "npx",
      "args": ["-y", "@modelcontextprotocol/server-git"]
    }
  }
}

Step 3: Test Connection

# List available tools
mcp list-tools

# Test single tool
mcp call-tool filesystem read_file --path /home/yourname/projects/README.md

Step 4: Start Chatting

Now you can interact with AI using natural language:

"Show me all Python files in the projects directory and count the lines of code"
"Find the 5 most recently modified files and generate a change summary"
"Add an installation section to README.md"

4. Best Practices and Security Recommendations

✅ Recommended Practices

Principle of Least Privilege: Only expose necessary directories and resources
Environment Variable Management: Store sensitive information in .env files
Audit Logging: Record all MCP tool calls
Version Pinning: Use fixed version numbers instead of latest

❌ Pitfalls to Avoid

Don’t expose root directory: / is off-limits
Don’t hardcode passwords: Use environment variables or secret management
Don’t trust all input: Validate file paths and query parameters
Don’t ignore error handling: MCP calls can fail

5. Advanced: Write Custom MCP Servers

If existing tools don’t meet your needs, write your own:

// Simplest MCP Server example
import { Server } from "@modelcontextprotocol/sdk/server";

const server = new Server({
  name: "my-custom-server",
  version: "1.0.0"
});

server.setRequestHandler("tools/call", async (request) => {
  if (request.params.name === "hello") {
    return {
      content: [{ type: "text", text: "Hello from MCP!" }]
    };
  }
});

server.listen();

Official SDKs:

TypeScript: https://github.com/modelcontextprotocol/typescript-sdk
Python: https://github.com/modelcontextprotocol/python-sdk

6. Summary and Future Outlook

The MCP ecosystem is evolving rapidly:

Tool Type	Maturity	Recommendation
File System	⭐⭐⭐⭐⭐	Must-Have
Database	⭐⭐⭐⭐	Highly Recommended
GitHub	⭐⭐⭐⭐	Developer Essential
Web Scraping	⭐⭐⭐	Use as Needed
Git	⭐⭐⭐⭐	Developer Essential