Nemlig Shopper 🛒

A CLI tool for shopping on Nemlig.com (Danish online grocery store). Parse recipes, search products, and add items to your cart.

Installation

# Install from PyPI
uv tool install nemlig-shopper

# Or run directly with uvx
uvx nemlig-shopper --help

Quick Start

# Log in to Nemlig.com
nemlig-shopper login

# Parse a recipe to see ingredients
nemlig-shopper parse "https://www.valdemarsro.dk/pasta-carbonara/"

# Search for products
nemlig-shopper search "mælk"

# Add a product to cart (by product ID from search results)
nemlig-shopper add 701015

# View your cart
nemlig-shopper cart

Commands

Command	Description
login	Authenticate with Nemlig.com
logout	Clear saved credentials
parse <url>	Parse recipe and display ingredient list
search <query>	Search Nemlig products
add <product_id>	Add product to cart
cart	View current cart contents

Usage Examples

Parse a Recipe

# From URL (supports 100+ recipe sites)
nemlig-shopper parse "https://www.valdemarsro.dk/lasagne/"

# From text input
nemlig-shopper parse --text "500g hakket oksekød
1 løg
2 fed hvidløg
400g hakkede tomater"

Search Products

# Basic search
nemlig-shopper search "økologisk mælk"

# Limit results
nemlig-shopper search "ost" --limit 5

Add to Cart

# Add single item
nemlig-shopper add 701015

# Add with quantity
nemlig-shopper add 701015 --quantity 2

Configuration

Credentials can be provided via:

1. Environment variables (.env file):

   NEMLIG_USERNAME=your-email@example.com
   NEMLIG_PASSWORD=your-password
   

2. Saved credentials: Run nemlig-shopper login to save credentials locally

Credentials are stored in ~/.nemlig-shopper/credentials.json (chmod 600).

Supported Recipe Sites

Uses recipe-scrapers supporting 100+ sites including:

• Valdemarsro (Danish)
• Mummum (Danish)
• AllRecipes
• BBC Good Food
• Serious Eats
• And many more...

Development

# Clone and install
git clone https://github.com/mhattingpete/nemlig-shopper.git
cd nemlig-shopper
uv sync

# Run tests
uv run pytest

# Run CLI locally
uv run nemlig --help

Using with Claude Code or LLM Agents

This CLI is designed to be agent-friendly. An LLM agent (Claude Code, custom agents, etc.) can drive the full shopping workflow by chaining CLI commands.

Prerequisites

1. Install the CLI (see Installation)
2. Set up credentials via environment variables or nemlig-shopper login
3. Point your agent at SKILL.md for the full command reference

Agent Workflow

A shopping list can contain a mix of recipe URLs and plain ingredients. The agent workflow is:

Shopping List (URLs + plain items)
  ├─ Recipe URLs → nemlig-shopper parse <url> → extract ingredients
  ├─ Plain items → use directly
  ↓
For each ingredient:
  → nemlig-shopper search "<danish ingredient name>" → get product IDs
  → nemlig-shopper add <product_id> --quantity <n> → add to cart
  ↓
nemlig-shopper cart → verify final cart

Example: Agent Shopping Session

# 1. Parse a recipe URL to get ingredients
nemlig-shopper parse "https://www.valdemarsro.dk/pasta-carbonara/"
# Output: list of ingredients with quantities and units

# 2. Search for each ingredient (Danish names work best)
nemlig-shopper search "spaghetti"
nemlig-shopper search "pancetta"
nemlig-shopper search "æg"
nemlig-shopper search "parmesan"

# 3. Add selected products by ID
nemlig-shopper add 701015 --quantity 1
nemlig-shopper add 503220 --quantity 1
nemlig-shopper add 100042 --quantity 1
nemlig-shopper add 504100 --quantity 1

# 4. Verify the cart
nemlig-shopper cart

Tips for Agent Integration

• Translation: Nemlig.com is Danish. Translate English ingredient names to Danish before searching (e.g., "milk" → "mælk", "onion" → "løg", "chicken" → "kylling").
• Product selection: Search results include product ID, name, price, size, and stock status. Pick products that are in stock and match the needed quantity/size.
• Quantities: The parse command outputs quantities and units per ingredient. Use these to determine how many units of a product to add.
• Multiple recipes: Process each recipe URL separately with parse, then search and add all ingredients.
• Plain items: Items like "mælk" or "rugbrød" that aren't from a recipe can be searched directly without parsing.
• Skill file: See SKILL.md for the complete agent-readable command reference.

MCP Server (Claude Desktop & Claude clients)

Instead of driving the CLI, you can run Nemlig Shopper as an MCP server so Claude searches, picks products, and fills your basket directly in chat. Paste a shopping list mixing recipe URLs and plain groceries and Claude handles the rest — it never places an order, it stops at a filled basket.

Setup

1. Install the package (see Installation) so the nemlig-mcp command is available.

2. Add the server to your client's MCP config. For Claude Desktop, edit claude_desktop_config.json:

   {
     "mcpServers": {
       "nemlig": {
         "command": "nemlig-mcp",
         "env": {
           "NEMLIG_USERNAME": "your-email@example.com",
           "NEMLIG_PASSWORD": "your-password"
         }
       }
     }
   }

   (With uvx: "command": "uvx", "args": ["--from", "nemlig-shopper", "nemlig-mcp"].)

3. Restart the client — Nemlig's tools then appear in chat.

Tools

Tool	What it does
search_products	Search Nemlig; returns candidates tagged cheapest / recommended / organic
pick_products	Like search_products, but renders an interactive picker (MCP Apps) so you click to add — falls back to the same list where rendering isn't supported
parse_recipe	Expand a recipe URL or ingredient text into structured ingredients
add_to_cart	Add a product (by id) to the basket
view_cart	Show the current basket and totals
clear_cart	Empty the basket

There is intentionally no checkout tool — review and place the order yourself at nemlig.com/basket.

Interactive picker: pick_products ships a clickable card UI via MCP Apps. It renders today in clients that support MCP Apps (e.g. claude.ai); Claude Desktop's support is still maturing, and there it degrades gracefully to the conversational candidate list.

Disabling it: set NEMLIG_MCP_APPS=0 (or false/no/off) in the server's env to turn the widget off entirely — pick_products and the UI resource aren't registered, leaving only the conversational tools.

Example prompt

Here's my list: https://www.valdemarsro.dk/pasta-carbonara/ plus milk, rye bread, and bananas. Find good options on Nemlig and fill my basket — ask me when you're unsure.

Notes

• Uses an unofficial Nemlig.com API
• Danish ingredient/product names work best
• Product IDs are shown in search results

License

MIT License - See LICENSE file for details