fooddb-jp 🇯🇵🍱

The complete Japanese Standard Tables of Food Composition (8th ed. 2023) as a structured SQLite database + REST API + MCP Server.

Japan's Ministry of Education (MEXT) provides food composition data only as Excel files. This project converts 2,541 foods × 353 nutrients (440,441 records) into an INFOODS Tagname-compliant SQLite database, queryable via API or AI agents (MCP).

🇯🇵 日本語 README はこちら

Quick Start

# Install dependencies
uv sync

# Start REST API (Swagger UI at http://localhost:8800/docs)
uv run uvicorn api:app --port 8800

# Start MCP Server (stdio mode)
uv run python mcp_server.py

🌐 Live API

A public API is available — no setup required:

# Search for chicken
curl "https://fooddb.navii.online/foods/search/chicken"

# Get vitamin C ranking (top 10)
curl "https://fooddb.navii.online/ranking/VITC?limit=10"

# Calculate nutrition (100g oatmeal + 200g milk)
curl "https://fooddb.navii.online/calculate?foods=1004:100,13003:200"

Base URL: https://fooddb.navii.online Swagger UI: https://fooddb.navii.online/docs Start free: https://fooddb.navii.online/start — copy a ready-to-paste keyless mcp_config.json and start on the Free tier (no API key).

Pricing

Free tier requires no API key — grab your MCP config at /start. For higher limits, visit /billing/checkout.

Database Scale

Why fooddb-jp?

Data Source

• Standard Tables of Food Composition in Japan (8th Revised Edition, Supplementary 2023)
• Published by: Ministry of Education, Culture, Sports, Science and Technology (MEXT)
• URL: https://www.mext.go.jp/a_menu/syokuhinseibun/mext_00001.html
• All 11 Excel files (Main table, Amino acids ×4, Fatty acids ×3, Carbohydrates ×3)

Data Quality

Automated quality checks (6 validations):

• ✅ No negative values in the entire dataset
• ✅ Mass balance (Water + Protein + Fat + Carbs + Ash ≈ 100g) — no anomalies
• ✅ Energy consistency — calculated vs. recorded values within tolerance
• ✅ Estimated value flags — 18.8% of records flagged as estimated (expected)
• ✅ Excel cross-check — 10 foods verified against original Excel (10/10 PASS)

REST API Endpoints

Performance (Mac Mini, SQLite)

MCP Tools (for AI Agents)

5 tools for direct AI agent integration:

Nutrient Identifiers (INFOODS Tagnames)

Key tags:

The trailing - is an official INFOODS suffix meaning "analytical method unspecified".

Project Structure

fooddb-jp/
├── data/
│   ├── raw/                     # Source data (MEXT Excel ×11)
│   └── output/                  # Converted intermediate files
├── scripts/
│   ├── convert_all.py           # Excel → SQLite pipeline
│   ├── build_nutrient_master.py # Nutrient definition builder
│   ├── build_aliases.py         # Search alias generator
│   ├── quality_check.py         # Data quality validation (6 checks)
│   └── test_food_name_parse.py  # Excel cross-validation test
├── fooddb.sqlite                # SQLite database (download from Releases)
├── api.py                       # FastAPI REST API
├── mcp_server.py                # MCP Server (stdio/HTTP)
└── pyproject.toml

License

• Code: MIT License
• Data: MEXT Government Standard Terms of Use 2.0 (equivalent to CC BY 4.0). Attribution: Ministry of Education, Culture, Sports, Science and Technology (MEXT), Japan.