Provides ChatGPT with tools to record, correct, and query meal nutrition, training sessions, and nutrition goals, including durable data storage and daily energy budget calculations.
Click on "Deploy Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type @ followed by the MCP server name and your instructions, e.g., "@mcp-nutrition-db Log my lunch: grilled chicken salad, quinoa, apple, and a glass of water."
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
mcp-nutrition-db is a private MCP service that lets ChatGPT record, correct,
and query meal nutrition, training sessions, and nutrition goals. ChatGPT
interprets meal photos and conversation context; this service owns the durable,
auditable data and calculates daily energy budgets.
The production service is intended to run on the NixOS host kage. It listens
only on loopback and is reached from ChatGPT through OpenAI Secure MCP Tunnel,
so it does not require an nginx route or a public MCP endpoint.
System design defines the product boundary, MCP tools, data model, security model, and NixOS deployment.
Implementation plan is the source of truth for phases, acceptance criteria, progress, and deferred work.
Exercise and recovery energy-credit policy defines confidence-adjusted training allowance, non-recurring recovery-day credits, bounded surplus repayment, recovery protection, and MCP presentation.
Local testing explains direct Codex attachment and the temporary Secure MCP Tunnel workflow for ChatGPT Web.
The implementation plan should be updated in the same commit as meaningful implementation milestones. Design changes should be recorded in the decision log before or alongside the code that depends on them.
Related MCP server: Food Facts MCP Server
The local MVP implements the SQLite schema and all sixteen MCP tools. Repository, schema, calendar, MCP process, Streamable HTTP, package, and NixOS evaluation checks pass; see the implementation plan for the exact verified state.
Enter the pinned environment and run the checks:
nix develop
PYTHONPATH=src ruff format --check src tests
PYTHONPATH=src ruff check src tests
PYTHONPATH=src mypy src
PYTHONPATH=src pytestRun a loopback development server with disposable state:
nix run . -- serve --database /tmp/mcp-nutrition-db.sqlite3Create an atomic SQLite online snapshot without stopping the server:
nix run . -- backup \
--database /tmp/mcp-nutrition-db.sqlite3 \
--output /tmp/mcp-nutrition-db.backup.sqlite3The MCP endpoint is http://127.0.0.1:8787/mcp; readiness is available at
http://127.0.0.1:8787/healthz. The server refuses a non-loopback HTTP bind by
default. See local testing to attach a new Codex session
or run the pinned tunnel client for ChatGPT Web.
This server cannot be deployed
Food logging, nutrition summaries, and meal photo calorie and macro estimates.
Log meals, check calories and macros, set up a nutrition plan, and search foods.
Personal nutrition tracking — log meals, track macros, review history, import from another app.
Log what you ate by talking to your AI assistant — calories and macros, completely free.
