A comprehensive collection of Model Context Protocol (MCP) tool patterns and recipes. This cookbook demonstrates various MCP design patterns that you can use as building blocks for your own AI-powered applications.
This cookbook demonstrates essential MCP tool patterns:
code_reviewer_prompt- Ready-to-use comprehensive code review promptsreact_prompt_generator- Dynamic ReAct (Reason + Act) pattern prompt generation- Use case: Transform MCP into a prompt repository service
smart_recipe_generator- Progressive information gathering with intelligent questions- Use case: Handle incomplete user input gracefully by asking targeted follow-up questions
step1_initialize_workflow→step2_execute_workflow- Sequential tool dependencies- Use case: Complex workflows requiring state management and ordered execution
sequential_web_search- Bridges external tools (like web_search) with MCP processing- Use case: Extend and orchestrate native client tools from the server side
tool_with_follow_up- Tools that suggest next actions through embedded commandsquantum_mood_analyzer,rubber_duck_debugger,coffee_brew_oracle- Fun example tools- Use case: Guide conversation flow through tool response content
See these MCP tool patterns demonstrated in real AI conversations:
The code_reviewer_prompt tool exposing a comprehensive, ready-to-use code review template
The smart_recipe_generator tool progressively gathering information through targeted questions
The sequential_web_search tool orchestrating multiple web searches and processing results
The tool_with_follow_up tool suggesting next actions, leading to rubber_duck_debugger execution
# Create virtual environment
python3 -m venv venv
# Activate virtual environment
source venv/bin/activate
# Install dependencies
pip install -r requirements.txtOr use the convenient activation script:
source activate.shConnect this cookbook to LLM applications that support MCP:
- Claude Desktop: Add to your MCP configuration
- Cursor: Add to your MCP configuration
- Custom LLM apps: Use the FastMCP client
- Other MCP clients: Any MCP-compatible client can use these patterns
{
"mcpServers": {
"mcp-tools-cookbook": {
"command": "python",
"args": ["path-to-repository/mcp_server.py"]
}
}
}Add to your ~/.cursor/mcp.json file:
{
"mcp-tools-patterns": {
"command": "python",
"args": [
"path-to-repository/mcp_server.py"
]
}
}This cookbook is provided as-is for educational purposes. Check the FastMCP repository for its license terms.