Raley
Automated grocery shopping assistant with intelligent unit pricing and automatic coupon clipping, enabling AI to search products, manage carts, and plan grocery runs via Claude Desktop or CLI.
README
Raley
Named after someone who made grocery day feel less like a hostage negotiation with your own pantry.
Automated grocery shopping assistant with intelligent unit pricing, automatic coupon clipping, and an MCP server so your AI can argue with produce departments on your behalf.
What This Does
You tell Raley what you need. Raley finds the best deal, calculates unit pricing across every bizarre measurement the grocery industry has invented (per oz, per lb, per "each" — what even is an "each"), clips relevant coupons, and manages your cart. All through a CLI or Claude Desktop.
The grocery store API uses F5 bot detection, so Raley shells out to curl for TLS fingerprint evasion instead of using Python HTTP libraries that get immediately flagged. Yes, we're using subprocess.run(["curl", ...]) in 2026. Sometimes the unglamorous solution is the one that actually works.
Install
curl -fsSL https://raw.githubusercontent.com/johnzfitch/raley-bot/main/install.sh | bash
This clones the repo, sets up a Python venv, installs dependencies, and symlinks raley-bot to ~/.local/bin. Requires git and curl (installs uv automatically if missing).
Then log in:
raley-bot login
Manual Install
git clone https://github.com/johnzfitch/raley-bot.git
cd raley-bot
uv venv && source .venv/bin/activate
uv pip install -e ".[login]"
raley-bot login
CLI
raley search "organic spinach" # Find products with unit pricing
raley history # Previously purchased items
raley offers # Available coupons
raley clip-all # Clip everything. All of it.
raley status # Session health check
raley orders # Order history
raley points # Something Extra balance
MCP Server (Claude Desktop / Claude Code)
Claude Desktop — add to ~/.config/Claude/claude_desktop_config.json (Linux) or ~/Library/Application Support/Claude/claude_desktop_config.json (macOS):
{
"mcpServers": {
"raley-bot": {
"command": "uv",
"args": ["run", "--directory", "/path/to/raley-bot", "raley-mcp"]
}
}
}
Claude Code — add to ~/.claude/settings.json:
{
"mcpServers": {
"raley-bot": {
"command": "/path/to/raley-bot/.venv/bin/raley-mcp",
"args": []
}
}
}
The install script configures these automatically.
Then talk to Claude like a normal person:
"Find me avocados, show me the best deal"
"Add organic spinach to my cart"
"What's in my cart?"
"Plan a grocery run: milk, eggs, bread, chicken thighs"
MCP Tools
| Tool | What It Does |
|---|---|
search |
Find products with unit pricing and value analysis |
add |
Add items to cart (returns product name for confirmation) |
remove |
Remove items by SKU |
cart |
View cart (supports summary_only, save_to_file, limit) |
offers |
List, clip all, or sync coupons to local DB |
plan |
Parse freeform grocery lists, find best matches (does not auto-add) |
price |
Check price history from local tracking DB |
orders |
Order history with date range and spend totals |
auth |
Session status (no cookie values exposed) |
Unit Pricing
Raley normalizes prices across every unit the grocery industry throws at you. A 6CT avocado bag at $4.99 becomes $0.83/unit. A 2lb chicken breast at $8.99 becomes $0.28/oz. Weight-based items show per-oz or per-lb depending on size. Liquids get per-ml. Count items get per-unit.
{
"sku": "10101877",
"name": "Hass Avocado, 6CT Bag",
"price": "$4.99",
"cents": 499,
"per_unit": "$0.83"
}
Architecture
raley_assistant/
├── api.py # Curl-based HTTP client (F5 evasion)
├── cli.py # Click CLI with Rich tables
├── mcp_server.py # MCP tools for Claude Desktop
├── unit_pricing.py # Normalize $/oz, $/lb, $/unit, $/ml
├── cart_builder.py # Freeform grocery list parsing
├── reasoning.py # Purchase frequency analysis
├── db.py # SQLite price history tracking
├── auth.py # Browser-based login (Helium/Selenium)
└── cookies.py # Session persistence
How the HTTP Client Works
The store's API sits behind F5 BIG-IP with browser fingerprinting. Python's requests and httpx get blocked because their TLS handshakes don't match real browsers. Rather than importing a Rust-based TLS emulation library (the previous approach using rnet, which created a phantom dependency that was never wired in), the client shells out to system curl which naturally produces a browser-like TLS fingerprint. Ugly, effective, zero exotic dependencies.
Data Storage
| What | Where |
|---|---|
| Session cookies | ~/.config/raley-assistant/cookies.json (mode 0600) |
| Price history DB | ~/.local/share/raley-assistant/raley.db |
| Saved results | ~/.local/share/raley-assistant/*.json (mode 0600) |
Security Notes
Session cookies are written with 0600 permissions (owner read/write only). Cookie values are sanitized for header injection characters before being passed to curl. No credentials, tokens, or cookie values are ever logged, committed, or included in MCP tool responses. The auth tool explicitly returns session status without exposing cookie contents.
The search tool is read-only — it does not silently clip coupons or modify cart state. Write operations (add, remove, offers clip_all) are always explicit.
Rate limiting is applied to bulk operations (200ms between requests, 500ms every 10th request) to avoid triggering bot detection.
Preferences
Copy the example and edit to taste:
cp preferences.example.json ~/.config/raley-assistant/preferences.json
Format:
{
"milk": {
"brand": "Clover",
"type": "whole",
"size": "gallon"
},
"general": {
"prefer_local": true,
"organic_preference": "indifferent",
"budget_target": 200
}
}
Troubleshooting
Session expired? → raley-bot login
Browser login not working? → Export cookies manually with a browser extension ("Cookie-Editor" or "EditThisCookie"), save as JSON, then raley login --file cookies.json
Database corrupted? → rm ~/.local/share/raley-assistant/raley.db (rebuilds on next search)
Coupon clip limit? → Some manufacturer coupons have daily claim limits. This is server-side, not a bug.
Development
uv pip install -e ".[dev]"
pytest
Disclaimer
This is an unofficial tool not affiliated with or endorsed by the grocery chain. Use responsibly and in accordance with their Terms of Service. This project is named after a person, not a brand.
License
Unlicense
Recommended Servers
playwright-mcp
A Model Context Protocol server that enables LLMs to interact with web pages through structured accessibility snapshots without requiring vision models or screenshots.
Magic Component Platform (MCP)
An AI-powered tool that generates modern UI components from natural language descriptions, integrating with popular IDEs to streamline UI development workflow.
Audiense Insights MCP Server
Enables interaction with Audiense Insights accounts via the Model Context Protocol, facilitating the extraction and analysis of marketing insights and audience data including demographics, behavior, and influencer engagement.
VeyraX MCP
Single MCP tool to connect all your favorite tools: Gmail, Calendar and 40 more.
graphlit-mcp-server
The Model Context Protocol (MCP) Server enables integration between MCP clients and the Graphlit service. Ingest anything from Slack to Gmail to podcast feeds, in addition to web crawling, into a Graphlit project - and then retrieve relevant contents from the MCP client.
Kagi MCP Server
An MCP server that integrates Kagi search capabilities with Claude AI, enabling Claude to perform real-time web searches when answering questions that require up-to-date information.
E2B
Using MCP to run code via e2b.
Neon Database
MCP server for interacting with Neon Management API and databases
Exa Search
A Model Context Protocol (MCP) server lets AI assistants like Claude use the Exa AI Search API for web searches. This setup allows AI models to get real-time web information in a safe and controlled way.
Qdrant Server
This repository is an example of how to create a MCP server for Qdrant, a vector search engine.