🛠 Writing Your First Tool¶
🛠 Writing Your First Tool¶
This tutorial walks you through creating a custom MCP tool.
Template available at: MCPStack-Tool-Builder
1. Use the MCPStack Tool Builder¶
Instead of writing everything by hand, you can bootstrap a tool skeleton:
git clone https://github.com/MCP-Pipeline/MCPStack-Tool-Builder.git
cd MCPStack-Tool-Builder
uv sync # or: pip install -e .[dev]
uv run mcpstack_tool.py init
Follow the interactive prompt to set:
tool_slug(e.g.weather)class_name(e.g.WeatherTool)env_prefix(e.g.WEATHER_)
Tip
Run uv run mcpstack_tool.py --help to see all available commands (preview, apply, validate, etc.).
2. Define your tool¶
Open src/mcpstack_<your_tool_name>/tool.py and add your actions:
from MCPStack.core.tool.base import BaseTool
class WeatherTool(BaseTool):
def actions(self):
return [self.get_weather]
def get_weather(self, city: str) -> str:
return f"Weather for {city}: Sunny ☀️"
3. Add CLI support¶
In cli.py implement a Typer app for your tool’s configuration:
import typer
from MCPStack.core.tool.base_cli import BaseToolCLI, ToolConfig
class WeatherCLI(BaseToolCLI):
@classmethod
def get_app(cls) -> typer.Typer:
app = typer.Typer()
@app.command()
def status():
print("Weather tool is configured.")
return app
@classmethod
def configure(cls) -> ToolConfig:
return {"env_vars": {}, "tool_params": {}}
4. Test your tool¶
Add it to a pipeline:
from MCPStack.stack import MCPStackCore
from weather_tool.tool import WeatherTool
stack = MCPStackCore().with_tool(WeatherTool())
stack.build(type="fastmcp")
stack.run()
Done!
You just wrote your first MCP tool with a CLI and action. Now stack it with other tools of interest!
