SafetySearch

Search. Scan. Stay Safe.

A comprehensive Model Context Protocol (MCP) server that provides access to FDA (Food and Drug Administration) data for Food safety information.

🎯 What This Server Provides

This MCP server offers 8 tools to access product safety data, helping users:

• Check product recalls and safety alerts for food products
• Monitor food safety issues and recall trends
• Analyze safety trends and company information
• Get comprehensive food safety insights

🚀 Quick Start

Prerequisites

• Python 3.10 or higher
• pip or uv package manager

Installation

1. Clone or download the project:

   git clone https://github.com/surabhya/SafetySearch.git
   cd SafetySearch
   

2. Install dependencies:

   # Using pip
   pip install "mcp[cli]>=1.0.0" httpx>=0.24.0 pydantic>=2.0.0
   
   # Or using uv (recommended)
   uv add "mcp[cli]>=1.0.0" httpx>=0.24.0 pydantic>=2.0.0
   

🔧 Usage

Using uv (Recommended)

Note: If you have previously installed mcp in another project, or if you encounter errors like Failed to spawn: mcp ... No such file or directory, run:

uv remove mcp
uv add "mcp[cli]>=1.0.0"

This ensures the mcp binary is correctly linked to your current environment.

Start the MCP Inspector (Development Mode)

Test and validate the server using the MCP Inspector:

uv run mcp dev server.py

Start the Server Directly

Run the server directly for testing:

uv run python server.py

Install in Claude Desktop (Production)

Install the server in Claude Desktop for production use:

uv run mcp install server.py

🛠️ Available Tools

Food Safety Tools (8 tools) ✅

Tool	Description	Parameters
search_recalls_by_product_description	Searches for food recalls with detailed analysis, safety insights, and comprehensive reporting.	query: str
search_recalls_by_product_type	Searches for recalls by product type with detailed analysis, company trends, and safety recommendations.	product_type: str
search_recalls_by_specific_product	Checks for recalls on specific products with detailed safety information and recommendations.	product_name: str
search_recalls_by_classification	Searches for recalls by classification with detailed analysis and risk assessment.	classification: str
search_recalls_by_code_info	Searches for recalls by code info with detailed product tracking and safety alerts.	code_info: str
search_recalls_by_date	Searches for recalls by date range with detailed timeline analysis and safety trends.	days: int (default: 30)
search_adverse_events_by_product	Searches for adverse events with detailed case analysis and safety insights.	product_name: str
get_symptom_summary_for_product	Gets detailed symptom analysis, case details, and safety insights for a specific food product.	product_name: str

🏛️ Architecture

The server is built with a simple, modular architecture designed for clarity and extensibility.

graph TD
    subgraph "SafetySearch MCP"
        A[User] -- "Tool Call" --> B["server.py<br/>(MCP Entrypoint)"];
        B -- "Executes" --> C{"Food Tools<br/>(safetyscore/tools/food.py)"};
        C -- "HTTP Request" --> D["API Client<br/>(safetyscore/api_client.py)"];
    end
    D -- "Calls" --> E["openFDA API<br/>(api.fda.gov)"];

    subgraph "Testing Framework"
      F[Pytest] -- "Runs" --> G["Test Suite<br/>(test_safetyscore/)"];
      G -- "Tests" --> C;
    end

• server.py: The main entry point of the MCP server. It initializes the toolsets and makes them available to the MCP environment.
• safetyscore/: The core Python package containing all the logic.
  • tools/: This directory contains the different tool modules. Currently, it only contains food.py.
    • food.py: Implements the 8 tools for food safety, which provide detailed analysis and safety insights.
  • api_client.py: A centralized asynchronous HTTP client for interacting with the external openFDA API. It handles request/response logic, error handling, and API key management.
• test_safetyscore/: Contains the test suite for the server, ensuring the reliability and correctness of the tools.

This structure separates concerns, making it easy to maintain and add new toolsets in the future.

📋 Example Prompts and Tool Calls

Here are some examples of user prompts and the corresponding tool calls they would trigger.

Food Safety Tools

• User Prompt: "Are there any recalls for ice cream?"

  food.search_recalls_by_product_description(query="ice cream")
  

• User Prompt: "Show me recent recalls for bakery products."

  food.search_recalls_by_product_type(product_type="Bakery")
  

• User Prompt: "I bought some 'Ben & Jerry's Chocolate Fudge Brownie' ice cream, is it safe?"

  food.search_recalls_by_specific_product(product_name="Ben & Jerry's Chocolate Fudge Brownie")
  

• User Prompt: "List all the most serious food recalls."

  food.search_recalls_by_classification(classification="Class I")
  

• User Prompt: "The code on my food package is '222268'. Is there a recall for it?"

  food.search_recalls_by_code_info(code_info="222268")
  

• User Prompt: "What food recalls have happened in the last two weeks?"

  food.search_recalls_by_date(days=14)
  

• User Prompt: "I got sick after eating Cheerios. Have other people reported problems?"

  food.search_adverse_events_by_product(product_name="Cheerios")
  

• User Prompt: "What are the common symptoms people report after eating 'Lucky Charms'?"

  food.get_symptom_summary_for_product(product_name="Lucky Charms")
  

🧪 Running Tests

To verify that all tools work as expected, you can run the provided test suites:

Prerequisites

• Ensure you have installed all dependencies (see Installation section above)

Run All Test Suites

From the project root directory, run:

# Test Food Tools
uv run python test_safetyscore/test_tools/test_food_tools.py

📊 API Endpoints Used

Food Safety

• Enforcement API: https://api.fda.gov/food/enforcement.json
• Adverse Events API: https://api.fda.gov/food/event.json

---

SafetySearch - Making FDA safety data accessible to everyone through the power of MCP.