SafetySearch is a modular MCP server that exposes FDA food safety data through eight specialized tools. It enables users to search recalls by product description, product type, specific product, classification, code information, and date range; and to examine adverse events and symptom summaries for foods. The server architecture centers on server.py as the MCP entrypoint, a safetyscore package housing tool logic, and an api_client that handles asynchronous HTTP requests to the openFDA API endpoints (enforcement and adverse events). The tools reside in safetyscore/tools/food.py and leverage openFDA data to provide detailed analyses, safety insights, and reporting. A test suite under test_safetyscore ensures reliability. The project relies on Python 3.10+, MCP CLI, and dependencies like httpx and pydantic. Endpoints used include https://api.fda.gov/food/enforcement.json and https://api.fda.gov/food/event.json. The Quick Start describes installation via MCP CLI or uv, and provides commands to run the inspector, start the server, and install for production. This description focuses on MCP-level access to FDA safety information, structured around modular tools.