🍩 NomNom Numbers

"Even donuts deserve to be counted."

Tell your AI what you ate. It remembers, tracks, and coaches — powered by real data, not vibes.

NomNom is a headless nutrition backend that plugs into Claude, OpenClaw, or any AI agent that can run a command. It searches 2M+ foods from the USDA, logs every meal to a local SQLite database, tracks your macros, builds streaks, and spots trends over time. Your AI handles the conversation. NomNom handles the truth.

You:    "I just had two scrambled eggs and toast for breakfast"

Claude: searches "scrambled eggs" → 140 cal, 12g protein
        logs "Scrambled Eggs" × 2 as breakfast
        checks today's progress

        "Logged! 🍩 You're at 850 / 2000 cal today with 65g protein.
         You've got room for a solid lunch — maybe that chicken bowl
         you liked on Thursday?"

Installation

# Run directly (recommended for AI agents)
bunx nomnomnumbers --help

# Or install globally
bun install -g nomnomnumbers
nomnom --help

Core Commands

Search & Log Meals

# Search 2M+ USDA foods
nomnom search "chicken breast" --limit 5

# Log by name (AI can search then log)
nomnom log "Chicken Breast" --qty 1 --calories 165 --protein 31

# Log by barcode
nomnom lookup 00000000924665
nomnom log "Quest Bar" --calories 200 --protein 21

# Save and reuse a repeatable meal
nomnom recipe create "Chicken Bowl" --calories 500 --protein 35 --carbs 42 --fat 18 --fiber 9
nomnom recipe log <recipe-id> --multiplier 2

Track Progress

# Today's summary
nomnom today --human

# Set goals
nomnom goals --calories 2000 --protein 150

# Progress vs goals (includes streaks!)
nomnom progress --human

History & Trends

# Recent meals
nomnom history --limit 10 --human

# Weekly trends (AI uses this for insights)
nomnom trends --days 7

# Find repeated meal combos worth saving as recipes
nomnom trends suggest-recipes --days 30 --min-occurrences 3

# Convert a suggestion into a saved recipe in one step
nomnom trends apply-suggestion <suggestion-id> --days 30 --min-occurrences 3

AI Integration

MCP (Claude Desktop, etc.)

Add to claude_desktop_config.json:

{
  "mcpServers": {
    "nomnom": {
      "command": "bunx",
      "args": ["nomnomnumbers", "mcp"]
    }
  }
}

Then Claude can:

• Log meals you mention in conversation
• Track your progress over time
• Give insights based on your actual data

Example Claude prompt:

"Track my nutrition using the nomnom MCP tool. When I tell you what I ate, log it. When I ask how I'm doing, check my progress. Be proactive about keeping me on track."

CLI (Any AI Agent)

AI agents that can run shell commands can use NomNom directly:

# All output is JSON for easy parsing
result=$(nomnom today)

# Exit code 0 = success, 1 = error
# stdout = JSON result
# stderr = errors/progress (never mixed)

Output Format

| Stream | Content | |--------|---------| | stdout | JSON result (add --human for readable) | | stderr | Errors, progress, initialization | | Exit 0 | Success | | Exit 1 | Error (stderr has { "error": "..." }) |

All Commands

| Command | Description | |---------|-------------| | init | Initialize database (auto-runs) | | search <query> | Search USDA foods | | lookup <barcode> | Look up by barcode | | log <food> [options] | Log a meal | | delete <id> | Delete a meal | | edit <id> [options] | Edit a meal | | today | Today's summary | | history | Meal history | | trends | Nutrition trends, recipe suggestions, and suggestion application | | goals | Set/view goals | | progress | Progress vs goals | | foods add/list/delete | Manage custom foods | | recipe create/list/log/delete | Manage reusable recipe templates | | config | View/modify config | | mcp | Start MCP server |

Run nomnom help for full details.

Data Storage

| Platform | Data | |----------|------| | Linux/Mac | ~/.local/share/nomnom/ | | Windows | %LOCALAPPDATA%\nomnom\ |

Override with NOMNOM_DATA_DIR environment variable.

Example: Agentic Weight Loss Workflow

Day 1 - Setup:

You: "I want to lose weight. My goal is 1800 calories, 140g protein."
Claude: nomnom goals --calories 1800 --protein 140
        "Goals set! I'll track your meals and keep you accountable."

Day 5 - Tracking:

You: "Had a turkey sandwich for lunch"
Claude: nomnom search "turkey sandwich"
        nomnom log "Turkey Sandwich" --calories 350 --protein 25
        "Logged! You're at 1200/1800 cal, 95/140g protein. 600 cal to go!"

Day 12 - Check-in:

You: "How am I doing?"
Claude: nomnom trends --days 7
        nomnom progress
        "Great week! Avg 1750 cal/day, hitting protein 6/7 days. 
         7-day streak on calories! Keep it up."

Day 30 - Insights:

You: "What patterns do you see?"
Claude: [Analyzes 30 days of data via nomnom history/trends]
        "You hit protein goals 85% of days you ate breakfast.
         On days you missed breakfast, avg calories were 2100 (over goal).
         Recommendation: Don't skip breakfast."

License

MIT