Custom Tool Implementation for MCP Server: 5-Minute Guide to Adding Unique Features¶

Target Audience

Intermediate developers with Python basics who want to leverage custom tools in Claude Code

Key Points¶

Add custom tools to MCP server
Execute those tools from Claude Code
Build foundation for practical automation tools

Why This Matters Now¶

MCP is the core of Claude Code extensibility, but practical implementation examples beyond standard tools are scarce, making custom business workflow automation unclear.

Solution Steps Overview¶

Step	Content	Success Metric
1	Implement basic MCP structure	server.py operational
2	Add custom tool functions	Tool registration success
3	Test Claude Code connection	Actual tool execution

Step 1: Basic MCP Server Implementation¶

Create minimal MCP server and prepare foundation for custom tools.

# server.py
import asyncio
from mcp import Tool
from mcp.server import Server
from mcp.types import ToolResult

server = Server("custom-tools")

@server.list_tools()
async def list_tools():
    return [
        Tool(
            name="file_validator",
            description="File existence check and basic info retrieval",
            inputSchema={
                "type": "object",
                "properties": {
                    "filepath": {"type": "string"}
                },
                "required": ["filepath"]
            }
        )
    ]

Step 2: Custom Tool Function Implementation¶

Implement practical file validation tool and register it with MCP server.

import os
import stat
from datetime import datetime

@server.call_tool()
async def call_tool(name: str, arguments: dict):
    if name == "file_validator":
        filepath = arguments["filepath"]

        try:
            if not os.path.exists(filepath):
                return ToolResult(content=f"❌ File not found: {filepath}")

            file_stat = os.stat(filepath)
            size_mb = file_stat.st_size / 1024 / 1024
            modified = datetime.fromtimestamp(file_stat.st_mtime)

            permissions = stat.filemode(file_stat.st_mode)

            result = f"""✅ File validation result:
📄 Path: {filepath}
📊 Size: {size_mb:.2f} MB
🕐 Last modified: {modified}
🔒 Permissions: {permissions}"""

            return ToolResult(content=result)

        except Exception as e:
            return ToolResult(content=f"❌ Error: {str(e)}", isError=True)

if __name__ == "__main__":
    asyncio.run(server.run())

Step 3: Claude Code Connection Setup¶

Add server information to claude_desktop_config.json to make it available from Claude Code.

{
  "mcpServers": {
    "custom-tools": {
      "command": "python",
      "args": ["path/to/server.py"],
      "env": {}
    }
  }
}

Common Pitfalls and Solutions¶

Symptom	Cause	Immediate Fix
Tools not showing	Config file path error	Use absolute path
Execution errors	Python environment mismatch	Unify with virtual env
Permission errors	File access permissions	chmod +x server.py

Advanced Configuration (Optimization)

### Multiple Tools Support

tools = [
    Tool(name="file_validator", ...),
    Tool(name="log_analyzer", ...),
    Tool(name="config_generator", ...)
]

### Enhanced Error Handling

try:
    # Tool execution
except FileNotFoundError:
    return ToolResult(content="File not detected", isError=True)
except PermissionError:
    return ToolResult(content="Insufficient access rights", isError=True)