---
description:
globs:
alwaysApply: false
---
---
description: MCP server development patterns and best practices
alwaysApply: false
---
# MCP Server Development
## Server Configuration Pattern
```python
config = ServerConfig(
name="MyServer",
description="Description of what this server does",
host="localhost",
port=8001,
transport="streamable-http", # or "stdio"
path="/mcp",
config={"type": "my_server", "custom_param": "value"}
)
```
## Tool Implementation Pattern
```python
@mcp.tool()
async def my_tool(param: str) -> str:
"""Tool description for AI to understand usage.
Args:
param: Description of the parameter
Returns:
Description of the return value
"""
# Validate inputs
if not param:
raise ValueError("Parameter cannot be empty")
# Implement tool logic
result = await some_async_operation(param)
return result
```
## Resource Implementation Pattern
```python
@mcp.resource("my-resource://path/{id}")
async def get_resource(id: str) -> str:
"""Resource description for AI context.
Args:
id: Resource identifier
Returns:
Resource content
"""
# Validate resource ID
# Return resource content
return "content"
```
## Best Practices
- All servers inherit from `BaseServer` and use `ServerConfig`
- Use `@mcp.tool()` decorator for tools, `@mcp.resource()` for resources
- Support both `stdio` and `streamable-http` transports
- Validate all inputs from AI models
- Handle errors gracefully with meaningful messages
- Use clear descriptions for AI consumption
- Log security-relevant operations
