New mcp tool
An operating system for Claude Code. Skills, hooks (MCR), and a lean-context framework for smarter AI coding sessions.
npx -y skills add justnau1020/claude-os --skill new-mcp-toolAssembled from the repository path, not quoted from the project. Check it against their README if it does not work.
One thing to look at
- 3 stars3 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.
What its author says it does
Copied from the file, not written here
Add a new MCP tool to a FastMCP server. Use when adding a new tool for LLM agents or extending the MCP interface with new functionality.
SKILL.md
1.7 KB, as published. Nobody here has run it
Add a New MCP Tool
Add a tool to the project's FastMCP server.
Arguments
$ARGUMENTS-- Tool name in snake_case (e.g.,get_user_profile)
Steps
1. Read the existing MCP server
Read the MCP server file to understand current tool patterns and the FastMCP setup.
2. Add the tool
@mcp.tool()
async def $ARGUMENTS(
param1: str,
param2: int = 10,
) -> str:
"""Clear description of what this tool does and when an LLM should use it.
Include example parameter values to help LLMs understand expected input.
Example: param1="example_value", param2=30
"""
...
3. Tool description conventions
The tool description is critical -- it's how LLMs decide when to call it:
- First sentence: what the tool does
- Second sentence: when to use it
- Include parameter examples inline
- Mention any constraints (rate limits, data availability)
4. Implementation conventions
- Async only: All tool functions must be
async def - Return strings: MCP tools return string responses for LLM consumption
- Actionable errors: If something fails, return a helpful error string, not a generic one
- Respect data boundaries: All data queries should respect the project's access control and data validation rules
5. Write tests
Add tests for the new tool's logic. Test:
- Happy path with valid parameters
- Error handling with invalid parameters
- Data validation on returned data
6. Verify
pytest tests/ -v -k "mcp"