Page8/8
Real-World MCP Applications Β· Page 1 of 1
MCP in Practice
Real-World MCP Applications
Database Server
MCP server for database access:
class DatabaseMCPServer(MCPServer):
def __init__(self, connection_string):
self.db = connect(connection_string)
@tool()
def query(self, sql: str) -> dict:
"Execute SQL query (read-only)"
return self.db.execute(sql)
@tool()
def describe_table(self, table: str) -> dict:
"Get table schema"
return self.db.describe(table)
# LLM can now query any database!
llm.chat("Show me top users by revenue")
# LLM calls: describe_table, query
# Returns: Results with proper SQL
Search Server
Multi-source search:
class SearchMCPServer(MCPServer):
@tool()
def web_search(self, query: str):
"Search the web"
return google_search(query)
@tool()
def code_search(self, query: str):
"Search GitHub"
return github_search(query)
@tool()
def academic_search(self, query: str):
"Search academic papers"
return arxiv_search(query)
# Single interface to all search sources!
Integration Server
Connect multiple APIs:
class IntegrationServer(MCPServer):
@tool()
def slack_send_message(self, channel: str, text: str):
"Send Slack message"
return slack.send(channel, text)
@tool()
def github_create_issue(self, repo: str, title: str):
"Create GitHub issue"
return github.create_issue(repo, title)
@tool()
def salesforce_update_lead(self, lead_id: str, data: dict):
"Update Salesforce lead"
return salesforce.update(lead_id, data)
# Agents can orchestrate across tools!
Custom Logic Server
Domain-specific business logic:
class BusinessLogicServer(MCPServer):
@tool()
def calculate_invoice(self, items: list):
"Calculate invoice with tax, discounts"
subtotal = sum(i["price"] for i in items)
discount = subtotal * 0.1 if subtotal > 1000 else 0
tax = (subtotal - discount) * 0.08
total = subtotal - discount + tax
return {
"subtotal": subtotal,
"discount": discount,
"tax": tax,
"total": total
}
@tool()
def check_inventory(self, product: str) -> bool:
"Check if in stock"
return inventory.get(product, 0) > 0
Why MCP Works
Before: Build integration for each LLM/framework
After: Build MCP server once, use everywhere
Database Server works with:
- Claude
- GPT-4
- LangChain agents
- Custom applications
- Future LLMs
Write once, deploy everywhere!
main.py
Loading...
OUTPUT
βΆClick "Run Code" to executeβ¦