If you're building AI agents with Claude, GPT, or any LLM, you've hit the same wall: your agent can't interact with the real world.
It can't check Bitcoin's current price. It can't look up where an IP address is located. It can't take a screenshot of a website. It can't run code.
MCP (Model Context Protocol) fixes this by giving your AI agent tools it can call. In this tutorial, I'll show you how to connect an MCP server that gives your agent 13 real-world tools — in under 5 minutes.
What You'll Build
An AI agent that can:
- Fetch live crypto prices (BTC, ETH, 1000+ tokens)
- Geolocate any IP address
- Look up DNS records
- Take website screenshots
- Execute Python/JavaScript code
- And more
Step 1: Get an API Key
curl -X POST https://agent-gateway-kappa.vercel.app/api/keys/create
Response:
{
"key": "fb_abc123...",
"credits": 200,
"message": "API key created. 200 free credits included."
}
Save that key — you'll need it.
Step 2: Configure the MCP Server
Add this to your Claude Desktop claude_desktop_config.json (or any MCP-compatible client):
{
"mcpServers": {
"frostbyte": {
"command": "npx",
"args": ["-y", "frostbyte-mcp"],
"env": {
"FROSTBYTE_API_KEY": "fb_abc123..."
}
}
}
}
That's it. Restart your client and your agent now has 13 tools.
Step 3: Try It
Ask your AI agent:
"What's the current price of Bitcoin and Ethereum?"
The agent will call the get_crypto_prices tool and return live data:
Bitcoin: $67,432.15
Ethereum: $3,891.22
Try these prompts:
"Where is the IP address 8.8.8.8 located?"
IP: 8.8.8.8
Country: United States
City: Mountain View, CA
ISP: Google LLC
Coordinates: 37.386, -122.084
"Take a screenshot of news.ycombinator.com"
The agent captures a real screenshot and returns it.
"Run this Python code: print(sum(range(1, 101)))"
Output: 5050
Exit code: 0
What Tools Are Available?
| Tool | Description |
|---|---|
get_crypto_prices |
Live prices for 1000+ tokens |
get_crypto_price |
Detailed price for a single token |
get_ip_geolocation |
Geolocate any IP address |
get_my_ip |
Get the agent's own IP |
dns_lookup |
Look up DNS records (A, MX, CNAME, etc.) |
take_screenshot |
Capture a website screenshot |
execute_code |
Run Python, JavaScript, TypeScript, or Bash |
get_wallet_balance |
Check crypto wallet balances (9 chains) |
get_defi_prices |
DeFi token prices and market data |
whois_lookup |
Domain WHOIS information |
check_ssl |
SSL certificate details |
port_scan |
Check if ports are open on a host |
web_scrape |
Extract content from web pages |
Using It Without MCP (Direct API)
Don't want MCP? Call the API directly:
# Crypto prices
curl https://agent-gateway-kappa.vercel.app/prices
# IP geolocation
curl "https://agent-gateway-kappa.vercel.app/api/geo/1.1.1.1" \
-H "X-API-Key: fb_abc123..."
# DNS lookup
curl "https://agent-gateway-kappa.vercel.app/api/resolve/github.com" \
-H "X-API-Key: fb_abc123..."
# Take a screenshot
curl "https://agent-gateway-kappa.vercel.app/api/screenshot?url=https://example.com" \
-H "X-API-Key: fb_abc123..."
# Run code
curl -X POST "https://agent-gateway-kappa.vercel.app/api/execute" \
-H "X-API-Key: fb_abc123..." \
-H "Content-Type: application/json" \
-d '{"language": "python", "code": "print(2**100)"}'
Why This Matters
Most AI agents today are blind — they can only work with data in their context window. By connecting real-world tools via MCP, you turn a chatbot into an agent that can:
- Monitor crypto portfolios with live data
- Investigate suspicious IP addresses
- Automate website monitoring with screenshots
- Debug code by actually running it
- Resolve DNS issues in real time
The MCP protocol is an open standard, so these tools work with Claude Desktop, Cursor, Windsurf, and any MCP-compatible client.
Getting Started
- Create a free API key (200 credits, no signup)
- Add the MCP server config to your client
- Start asking your agent to do real things
curl -X POST https://agent-gateway-kappa.vercel.app/api/keys/create
The full API docs are at agent-gateway-kappa.vercel.app.
Built with Node.js, Fastify, and a lot of caffeine. The MCP server source is on GitHub.
Top comments (0)