Skip to main content

Getting Started with MCP: A Complete Beginner's Guide

· 8 min read
ToolBoost Team
ToolBoost Team
ToolBoost Engineering Team

Ready to supercharge your AI workflows with Model Context Protocol (MCP)? This step-by-step guide will take you from zero to productive in under 30 minutes.

By the end of this tutorial, you'll have:

  • ✅ Deployed your first MCP server
  • ✅ Connected it to your AI client
  • ✅ Used real tools to accomplish actual tasks

Let's dive in!

What You'll Need​

Before we start, make sure you have:

  1. An AI Client - At least one of:

  2. A ToolBoost Account (free tier is perfect)

  3. 5-10 Minutes of your time

That's it! No coding required, no infrastructure to set up.

Step 1: Create Your ToolBoost Account (2 minutes)​

1.1 Sign Up​

Go to toolboost.dev/sign-up and create an account:

  • Enter your email
  • Create a password
  • Click "Create Account"

1.2 Verify Email​

Check your inbox and click the verification link. This activates your account.

1.3 Welcome!​

You'll be taken to your dashboard. Welcome aboard! 🎉

Step 2: Create Your First Project (2 minutes)​

Projects organize your MCP deployments. Let's create one:

2.1 Click "Create Project"​

On your dashboard, find the "Create Project" button (or navigate to Dashboard → Projects).

2.2 Name Your Project​

Name: My First MCP Project
Description: Learning to use MCPs with ToolBoost

Click "Create".

2.3 Save Your API Key​

IMPORTANT: You'll see your API key displayed. This is shown only once.

Your API Key: tb_abc123def456...

Copy this and save it somewhere safe (like a password manager). You'll need it to connect your AI clients.

tip

If you lose your API key, you can regenerate it from project settings. But you'll need to update all your client configurations.

Step 3: Deploy Your First MCP (3 minutes)​

Let's deploy the Filesystem MCP - it's simple and useful for beginners.

3.1 Browse the Catalog​

Click "Catalog" in the navigation menu. You'll see thousands of MCP servers.

3.2 Find Filesystem MCP​

Use the search bar:

Search: "filesystem"

Click on the "Filesystem MCP" card (usually by modelcontextprotocol).

3.3 Learn About It​

On the MCP detail page, you'll see:

  • Overview: What it does
  • Tools: Available functions (read_file, write_file, list_directory, etc.)
  • Environment Variables: Configuration needed

3.4 Deploy It​

  1. Select your project from the dropdown in the sidebar
  2. Click "Deploy Server"
  3. Configure environment variables:
ALLOWED_DIRECTORIES: /Users/your-username/Documents

(Change this to a directory you want the MCP to access)

  1. Click "Attach & Save Config"

🎉 Success! Your first MCP is deployed!

Step 4: Get Your Connection URL (1 minute)​

Now you need the URL to connect your AI client.

4.1 Find the "Use" Section​

On the MCP detail page, scroll to the "Use" section in the right sidebar.

4.2 Copy Your Connection URL​

You'll see:

Connection URL:
https://toolboost.dev/server/modelcontextprotocol/servers-filesystem/mcp?api_key=YOUR_API_KEY

Click "Copy URL" to copy it to your clipboard.

Step 5: Connect to Your AI Client (5 minutes)​

Now let's connect your deployed MCP to your AI client. Choose your client below:

Option A: Claude Desktop​

1. Find the Config File​

macOS:

~/Library/Application Support/Claude/claude_desktop_config.json

Windows:

%APPDATA%\Claude\claude_desktop_config.json

2. Edit the File​

Open it in a text editor. Add this configuration:

{
  "mcpServers": {
    "filesystem": {
      "serverUrl": "PASTE_YOUR_CONNECTION_URL_HERE"
    }
  }
}

Replace PASTE_YOUR_CONNECTION_URL_HERE with the URL you copied.

3. Restart Claude Desktop​

Completely quit Claude Desktop and restart it.

Option B: Cursor​

1. Open Settings​

Press Cmd/Ctrl + , to open settings.

2. Edit settings.json​

Click "Open Settings (JSON)".

3. Add Configuration​

{
  "claude-code.mcpServers": {
    "filesystem": {
      "serverUrl": "PASTE_YOUR_CONNECTION_URL_HERE"
    }
  }
}

4. Reload Window​

Press Cmd/Ctrl + Shift + P, type "Reload Window", and press Enter.

Option C: Windsurf​

1. Open Settings​

Go to Settings → Advanced → Cascade.

2. Edit mcp_config.json​

{
  "mcp": {
    "servers": {
      "filesystem": {
        "url": "PASTE_YOUR_CONNECTION_URL_HERE",
        "transport": "http"
      }
    }
  }
}

3. Restart Windsurf​

Quit and restart Windsurf.

Step 6: Test Your MCP (5 minutes)​

Time to see it in action!

6.1 Open Your AI Client​

Open Claude Desktop, Cursor, or Windsurf.

6.2 Try Basic Commands​

Let's test the Filesystem MCP:

Command 1: List Directory

Can you list the files in my Documents folder?

Expected Result: Your AI should use the Filesystem MCP to list files and show you the results!

Command 2: Read a File

Read the contents of [filename] in my Documents folder

(Replace [filename] with an actual file from the previous list)

Expected Result: The AI reads and displays the file contents.

Command 3: Create a File

Create a new file called "test.txt" in my Documents folder with the content "Hello from MCP!"

Expected Result: File created successfully!

6.3 Verify​

Check your Documents folder - you should see the test.txt file!

Step 7: Explore More MCPs (Optional)​

Now that you have one MCP working, let's add more!

1. Memory MCP​

Gives your AI persistent memory across conversations.

Deploy: Memory MCP
Use case: "Remember that my favorite color is blue"
         Later: "What's my favorite color?"

2. Time MCP​

Provides current date/time information.

Deploy: Time MCP
Use case: "What's the current date and time?"
         "What day of the week is Christmas this year?"

3. GitHub MCP​

Interact with GitHub repositories.

Deploy: GitHub MCP
Configure: Add GitHub Personal Access Token
Use case: "List my GitHub repositories"
          "Show issues in my react-app repo"

How to Add More:​

  1. Go to toolboost.dev/catalog
  2. Find an MCP you want
  3. Click "Deploy Server"
  4. Copy the connection URL
  5. Add to your AI client config:
{
  "mcpServers": {
    "filesystem": {
      "serverUrl": "..."
    },
    "memory": {
      "serverUrl": "PASTE_MEMORY_MCP_URL_HERE"
    },
    "time": {
      "serverUrl": "PASTE_TIME_MCP_URL_HERE"
    }
  }
}

Troubleshooting​

MCP Not Showing Up​

Problem: AI client doesn't recognize your MCP.

Solutions:

  • ✅ Verify you completely restarted your AI client
  • ✅ Check JSON syntax (no trailing commas!)
  • ✅ Ensure connection URL is correct (copy fresh from ToolBoost)
  • ✅ Verify MCP is deployed in your ToolBoost project

Authentication Errors​

Problem: "Unauthorized" or "Invalid API key" errors.

Solutions:

  • ✅ Copy the entire connection URL (including ?api_key=...)
  • ✅ Check for typos in the URL
  • ✅ Verify your API key hasn't been regenerated
  • ✅ Ensure you're using the correct project

MCP Not Working​

Problem: MCP connects but doesn't work properly.

Solutions:

  • ✅ Check environment variables are configured correctly
  • ✅ Verify required permissions (file paths, API tokens, etc.)
  • ✅ Try using the Toolspector on ToolBoost to test directly
  • ✅ Check ToolBoost status: status.toolboost.dev

Understanding What Just Happened​

Let's break down the architecture:

┌─────────────────────┐
│   Your AI Client    │ <-- You interact here
│  (Claude/Cursor)    │
└──────────┬──────────┘
           │
           │ HTTP/SSE
           │ (Connection URL)
           │
┌──────────┴──────────┐
│      ToolBoost      │ <-- Hosts your MCP
│    (Cloud Server)   │
└──────────┬──────────┘
           │
           │ Executes
           │
┌──────────┴──────────┐
│   Filesystem MCP    │ <-- Performs actions
│      (Tools)        │
└──────────┬──────────┘
           │
           ▼
┌─────────────────────┐
│   Your Computer     │ <-- Accesses your files
│      (Files)        │
└─────────────────────┘

Key points:

  1. Your AI client sends requests to ToolBoost
  2. ToolBoost runs the MCP server in the cloud
  3. The MCP executes the requested tool
  4. Results flow back to your AI client

Security: Your API key authenticates requests. Environment variables are encrypted.

Next Steps​

Congratulations! You've successfully:

  • ✅ Created a ToolBoost account
  • ✅ Deployed your first MCP
  • ✅ Connected it to your AI client
  • ✅ Used real MCP tools

Where to Go from Here:​

1. Explore More MCPs​

Browse the catalog and try:

  • Database MCPs (PostgreSQL, Supabase)
  • Development MCPs (GitHub, Docker)
  • Productivity MCPs (Web Search, Email)

2. Learn Advanced Features​

  • Configure environment variables
  • Manage multiple projects
  • Monitor usage and analytics
  • Invite team members

3. Read Use Cases​

See 10 real-world MCP use cases for inspiration.

4. Join the Community​

  • Share your workflows
  • Ask questions
  • Suggest new MCPs
  • Report issues

Quick Reference​

Your ToolBoost URLs​

Common Commands​

List files:

Show me the files in my Documents folder

Read file:

Read the contents of myfile.txt

Create file:

Create a file called notes.txt with this content: [your content]

Get help:

What tools are available in the filesystem MCP?

Free Tier Limits​

Your free ToolBoost account includes:

  • ✅ 10 deployed MCPs
  • ✅ 10,000 API requests per month
  • ✅ Access to entire MCP catalog
  • ✅ Email support
  • ✅ All core features

Need more? Check out our pricing.

Get Help​

Stuck? We're here to help:

Conclusion​

You're now part of the MCP revolution! With ToolBoost, you can:

  • Connect your AI to real tools
  • Automate workflows
  • Access external data
  • Build powerful integrations

And you did it all in under 30 minutes, with zero infrastructure setup.

What will you build next?

Share your creations with us at contact@toolboost.dev - we'd love to see what you're building!

Happy building! 🚀

New to ToolBoost? Sign up free and deploy your first MCP in minutes.