mcpField
A
Arugula

Arugula

Remote
arugulanet.comSearch0Updated 3d ago

About Arugula

Arugula is an offer-matching service for shopping requests, available through MCP clients, chatbots, and website search. It ranks merchant offers by semantic fit and returns prices, brief match explanations, and purchase links.

Key features

  1. 01Match offers by request meaning
  2. 02Rank results by fit only
  3. 03Show price, fit reason, and purchase link
  4. 04Return no results when nothing fits

Use cases

  1. 01Find suitable products through an MCP client
  2. 02Display relevant offers with website search
  3. 03Return matching merchant offers from a chatbot

MCP support

Arugula provides a Streamable HTTP MCP server at https://mcp.arugulanet.com/mcp, with no key or sign-in required. Its tools find offers, retrieve results from a search, and explain why an offer matches.

Configuration

Remote servers need no local install. Pick your client, then copy the config or install in one click.

Endpointhttps://mcp.arugulanet.com/mcp?ch=mcp-registry
Copy install command
Claude CodeTerminal
Cursor.cursor/mcp.json
VS Code.vscode/mcp.json
Claude Desktopclaude_desktop_config.json
Windsurf~/.codeium/windsurf/mcp_config.json

Run commands in a terminal; paste JSON into the client’s MCP config and restart. Servers that need sign-in will prompt on first connect.

Tools

Details

From the project README

Arugula, Intent Exchange Network

Arugula

Offers that fit, wherever people ask.

Arugula is a matching engine. Someone says what they want, in their own words, in their assistant, on a website or in a chat. Arugula matches the request on meaning against merchants' offers and returns the ones that fit, best first. Each offer comes with its price, a one-line reason why it fits, and a link to buy from the merchant.

  • Ranked by fit only. Nobody can pay to be placed higher.
  • Commission, disclosed. Arugula may earn a commission if you buy through these links, at no extra cost to you. Every list of offers says so.
  • No account needed. Every surface below works without signing up.
  • When nothing fits, nothing is shown.

Website: arugulanet.com · Privacy: arugulanet.com/privacy · Terms: arugulanet.com/terms

By using Arugula you agree to arugulanet.com/terms. Privacy: arugulanet.com/privacy.

Where to use it

SurfaceForStatus
MCP serverClaude, Cursor, Open WebUI, LibreChat, Jan, LM Studio and any MCP clientLive
TelegramChats and groupsLive
Site searchAny website with a search boxLive
ShopifyShopify storesLive
Google Tag ManagerSites that use GTMLive (import); gallery listing in review
WordPressWordPress sitesLive (ZIP); WordPress.org listing in review
Cloudflare ZarazSites that use ZarazLive
Cloudflare WorkerSites on Cloudflare without ZarazLive
ChatbotsChatwoot, Botpress, Rasa, Voiceflow and your own botLive, partner key on request
PythonYour own agents (LangChain, LangGraph, CrewAI)Live on PyPI
JavaScriptYour own agentsComing soon
Chrome and Firefox extensionsYour browserComing soon
Claude directory, ChatGPT appAssistantsComing soon
DiscordServersComing soon

MCP server

https://mcp.arugulanet.com/mcp: Streamable HTTP. No key and no sign-in.

ToolWhat it does
find_offersTakes a request in plain words and returns the offers that fit, best first.
get_offersGets the offers found for a request from find_offers; matching can take a few seconds.
explain_matchSays why an offer fits a request.

Each client below has its own address, ending in ?ch=. That tag only tells us which kind of client sent a request, so we can see which ones people use.

Claude (web and desktop): Settings, then Connectors, then Add custom connector, with https://mcp.arugulanet.com/mcp?ch=claude-connector. Custom connectors are available on some Claude plans only.

Claude Code:

claude mcp add --transport http arugula "https://mcp.arugulanet.com/mcp?ch=claude-connector"

Cursor: add this to ~/.cursor/mcp.json (every project) or .cursor/mcp.json (one project). This repository is also a Cursor plugin (.cursor-plugin/plugin.json and mcp.json).

{
  "mcpServers": {
    "arugula": {
      "url": "https://mcp.arugulanet.com/mcp?ch=cursor"
    }
  }
}

One-click install, if your system opens cursor:// links:

cursor://anysphere.cursor-deeplink/mcp/install?name=arugula&config=eyJ1cmwiOiJodHRwczovL21jcC5hcnVndWxhbmV0LmNvbS9tY3A/Y2g9Y3Vyc29yIn0=

Open WebUI (0.6.31 or later): Admin Settings, then External Tools, then + Add Server. Type MCP (Streamable HTTP), URL https://mcp.arugulanet.com/mcp?ch=open-webui, Auth None. The URL must be reachable from the Open WebUI server.

LibreChat: in librechat.yaml:

mcpServers:
  arugula:
    type: streamable-http
    url: https://mcp.arugulanet.com/mcp?ch=librechat
    title: Arugula
    description: Find offers that fit what you want, ranked by fit only.
    timeout: 60000

Jan: Settings, then MCP Servers, then + Add MCP Server. Name arugula, transport HTTP, URL https://mcp.arugulanet.com/mcp?ch=jan. Turn tools on for the model you use.

LM Studio (0.3.17 or later): Program tab, then Install, then Edit mcp.json, and add this inside mcpServers:

"arugula": {
  "url": "https://mcp.arugulanet.com/mcp?ch=lm-studio"
}

Any other MCP client: use https://mcp.arugulanet.com/mcp with the Streamable HTTP transport and no authentication.

The entry for the official MCP registry is server.json (com.arugulanet/arugula).

Telegram

Open @ArugulaNetBot and say what you want. To use it with others, add it to a group. To use it in any chat, type @ArugulaNetBot followed by your request.

Site search

One line on any page with a search box. When a visitor searches, offers that fit show up under your search form. The script sets no cookies.

<script async src="https://api.arugulanet.com/embed/v0.5.2/search.js"
        integrity="sha384-VUDq6tLbkK5ID6JJ8KnwaPN9dWB5VCKUxZ/r7gJTZdpXxn+yqTE2S7AKsVNmuXGD"
        crossorigin="anonymous" data-channel="site-search"></script>

It finds your search box by itself: input[type=search], a search form, or an input named q, s, search or query. All of these attributes are optional:

AttributeWhat it setsDefault
data-selectorYour search input, for example #qFound automatically
data-containerWhere the offers goRight after the search form
data-limitHow many offers to show3

The hashes for each release are at https://api.arugulanet.com/embed/v0.5.2/integrity.json. Each release path never changes.

Shopify

In your theme, open Edit code, then layout/theme.liquid, and paste this just before </body>:

<script async src="https://api.arugulanet.com/embed/v0.5.2/search.js"
        integrity="sha384-VUDq6tLbkK5ID6JJ8KnwaPN9dWB5VCKUxZ/r7gJTZdpXxn+yqTE2S7AKsVNmuXGD"
        crossorigin="anonymous" data-channel="shopify"></script>

It attaches to your store's search the same way as site search.

Google Tag Manager

The tag template is template.tpl in this repository. It loads the site-search script.

  1. In GTM, open Templates, then New under Tag Templates.
  2. Open the menu at the top right, choose Import, and select template.tpl. Save.
  3. Add a new tag of type Arugula that fires on All Pages. Publish the container.

You can also download the template as a ZIP from https://api.arugulanet.com/arugula-gtm-template.zip.

WordPress

Download arugula-wordpress.zip. In WordPress, open Plugins, then Add New, then Upload Plugin. Choose the file and activate it. It attaches to your site's search with nothing to set up. The plugin's readme.txt lists everything it sends and stores.

Cloudflare Zaraz and Workers

Three ways in, all in cloudflare/:

  1. Zaraz Custom HTML: add a Custom HTML tool with the site search snippet, using data-channel="zaraz", and give it a Pageview action.
  2. Zaraz Custom Managed Component: deploy cloudflare/component as a Worker with npx managed-component-to-cloudflare-worker ./index.js arugula, then add it under Third-party tools.
  3. Cloudflare Worker for sites on Cloudflare without Zaraz: cloudflare/worker adds the script at the edge (npx wrangler deploy).

The script runs once per page even when Zaraz fires on every single-page navigation.

Chatbots

A bot you already run can forward a user's message to Arugula and show the offers that come back:

POST https://api.arugulanet.com/v1/partner/intents
Authorization: Bearer <partner key>

{"text": "a linen sofa under S$1,500", "user_id": "<your user's id>"}

The answer has offers, each with a tracked_url, and a ready-made text reply. For Chatwoot, set the agent bot's webhook to https://api.arugulanet.com/hooks/chatwoot/<partner id>?key=<partner key>. To get a partner key, open an issue.

Python

arugulanet on PyPI. Needs Python 3.9 or later, has no dependencies, and needs no account.

pip install arugulanet
from arugulanet import Arugula

offers = Arugula().find_offers("a hotel room in Singapore for two nights", party_size=2)

For LangChain and LangGraph, install "arugulanet[langchain]" or "arugulanet[langgraph]" and use from arugulanet.langchain import find_offers_tool. For CrewAI, install "arugulanet[crewai]" and use from arugulanet.crewai import FindOffersTool.

JavaScript

Coming soon to npm as arugulanet. Needs Node 18 or later and has no runtime dependencies. Includes tools for OpenAI function calling, LangChain.js and CrewAI-style agents.

import { Arugula } from "arugulanet";

const offers = await new Arugula().findOffers("a hotel room in Singapore for two nights", { party_size: 2 });

Showing offers

If you build on Arugula, show this line with the offers:

Ranked by fit only. Arugula may earn a commission if you buy through these links, at no extra cost to you.

The tools and the API return it with every list of offers, as disclosure.

Brand

Logos, icons, avatars, favicons and social images are in brand/, with the rules in brand/BRAND.md. Use them as they are: no recolouring, outlines or effects.

Help

Questions, problems and requests: open an issue.

Licence

This repository is under the Apache License 2.0. The packages carry their own licences: MIT for the Python and JavaScript packages, and GPLv2 or later for the WordPress plugin.

Comments

Markdown supported

New to MCP?

From what the protocol is to installing your first server, these answers cover most first questions.

An open protocol that defines how AI applications connect to external tools and data in a uniform way. Clients talk to the model; MCP servers provide tools, resources and prompt templates. They exchange standard messages, so one server can be reused by many clients.