MCP App Store
by KeenEthics
TheBeanGeek icon

TheBeanGeek MCP Server for ChatGPT

By TheBeanGeek

Overview

TheBeanGeek is a specialty-coffee directory: browse coffees and roasteries with structured filters (no free-text search), see a roastery leaderboard, get personalized recommendations from your own review history, and log reviews for coffees - including ones not yet in the catalog, with an optional bag photo. The interactive workbench opens as a fullscreen app (My Coffee, Browse, and Review views) from the sidebar launcher or a conversation side panel, and roasteries can be @-mentioned in the composer. It only accesses your own TheBeanGeek account data.

How to connect

  1. ChatGPT: Open the TheBeanGeek listing in the ChatGPT app directory and choose Connect. ChatGPT listing TheBeanGeek exposes 20 tools here.

Tools

addReview

ChatGPT
Review a coffee already in the catalog. Collect the required fields from the user in ONE turn BEFORE any tool call — rating, brewMethod, tastes and place — then resolve the coffee once via searchCoffees (take its slug) and submit a single addReview call. A full review requires rating, brewMethod, tastes and place — ask the user for any that are missing, never invent. tastes is the user's own recognized flavors (see that field), not the catalog notes; it may be the literal "none" when they recognized nothing. Proactively ask for roastDate (roast date on the bag) and note (a short custom review note) even though optional — omit them only if the user genuinely does not have them.

coffee.open

ChatGPT
Open the TheBeanGeek app: the signed-in user's reviews and recommendations, plus catalog browsing and coffee review logging. Use when the user asks to open their coffee app, see their coffee profile, browse coffees visually, or review a coffee from the app. Optionally pre-filter the Browse view (view: 'catalog') with tastes, brewRoast, beanCountries, coffeeType, or isDecaffeinated.

coffee.panel

ChatGPT
Open the coffee panel beside this conversation.

discoverCoffees

ChatGPT
Recommend coffees from the user's review-derived taste profile. Needs several reviews — on PROFILE_NOT_READY do not retry: ask what they like (tastes, roast, brew method, origins) and searchCoffees with those filters instead. Every addReview sharpens future recommendations.

followRoastery

ChatGPT
Follow a roastery for the user (keep them posted on its new coffees; same follow as on the website). Only when the user asks to follow / favorite a roastery. Already following -> changed: false, treat as success.

getMyFollowedRoasteries

ChatGPT
Roasteries the user follows ("keep me posted" on new drops; the site may call it favorite), most recently followed first. Same roastery shape as searchRoasteries plus followedAt. Follows made on the website show here too. Up to 50 per call; page with offset.

getMyReviews

ChatGPT
The token owner's reviews, newest first: rating, brew method, tastes, place, roast date, note and the coffee snapshot (name, url, tastes, origins, roastery).

getMySavedCoffees

ChatGPT
Coffees the user saved to try later (their wishlist), most recently saved first. Same coffee shape as searchCoffees plus savedAt; check isUnavailable before suggesting a purchase. Saves from the website and the weekly-match email show here too. When presenting a coffee, ALWAYS show its photo: paste imageMarkdown, then the coffee url link below it. Up to 50 per call; page with offset.

reviewNonExistingCoffee

ChatGPT
Review a coffee NOT in the catalog (e.g. from a bag photo or roaster page). This is user-generated content, not catalog data: the coffee and review are recorded on TheBeanGeek's own database to the signed-in account's private UGC queue and, once approved, the coffee joins the public catalog. No third-party system is contacted. Collect the required fields from the user in ONE turn BEFORE any tool call — name, rating, brewMethod and place — then resolve the roastery once via searchRoasteries (verify ownership; see its description) and submit a single reviewNonExistingCoffee call. Required: roasterySlug plus name, rating, brewMethod, place. Also proactively ask for roastDate (roast date on the bag) and note (a short custom review note) — omit them only if the user genuinely does not have them. All other fields are best-effort: never invent, omit what is unknown. imageUrl: if the user attached a photo, call uploadCoffeePhoto first and pass its url — the url is stored verbatim, the server never fetches it; never invent a url. On COFFEE_ALREADY_EXISTS, confirm the match with the user and use addReview instead.

roastery.mentions

ChatGPT
At-mention picker for roasteries. Called by the host composer, not the model.

saveCoffee

ChatGPT
Save a catalog coffee to the user's wishlist (want to try later; same list as on the website). Only when the user asks to save / bookmark / remember a coffee. Already saved -> changed: false, treat as success.

searchCoffees

ChatGPT
Search the catalog with structured filters — no free text; map the user's words to the enum values ("fruity Ethiopian" -> tastes: ["fruity"], beanCountries: ["ET"]; "espresso" -> brewing: "espresso"). Results carry full coffee details: taste + tasteDetails, roasting, brewRoast, decaf/blend/unavailable flags, prices, rating, reviewCount, beans (origin, farm, varieties, processing) and roastery (slug, address, country, shipsTo, discount). Up to 10 per call; page with offset.

searchRoasteries

ChatGPT
Search roasteries by partial name and/or country. Results carry name, slug, url, address, country, shipsTo, coffeeCount, rating, verified flag, description and discount code. Several roasteries share a name: before using a slug in a review, match the returned address/country against the bag label or product page — confirm with the user if unsure. Up to 10 per call; page with offset.

submitFeedback

ChatGPT
Report a problem or request a feature about this coffee agent (thebeangeek MCP). This is an internal feedback channel: the note is stored in TheBeanGeek's own feedback table for the quality loop of this very server, where the team reads and triages it. Choose the type that fits: bug (something is broken), feature_request (something you want), data_error (a coffee/roastery fact looks wrong), or general (anything else). content is the free-text body — include enough detail to act on.

topRoasteries

ChatGPT
Top 10 roasteries by recent engagement: each entry is the roastery (same shape as searchRoasteries) plus its current and previous engagement values and rank.

unfollowRoastery

ChatGPT
Stop following a roastery for the user. Only when the user asks. Not following -> changed: false, treat as success.

unsaveCoffee

ChatGPT
Remove a coffee from the user's wishlist. Only when the user asks. Not saved -> changed: false, treat as success.

uploadCoffeePhoto

ChatGPT
Upload a bag photo so it can be attached to a review. Pass the photo as a base64 data URL (data:image/jpeg;base64,...) — the format image attachments are delivered as. Returns { url }; pass that url as imageUrl to reviewNonExistingCoffee. Only jpeg/png/webp up to 4 MB.

workbench.catalog

ChatGPT
Catalog list for the Browse view. Called by the app, not the model.

workbench.data

ChatGPT
Data for the My Coffee view. Called by the app, not the model.

Capabilities

InteractiveWrites

Example Prompts

Click any prompt to copy it.

Similar Apps

Listed on MCP App Store

Listed on MCP App Store

Add this badge to your site or README. It links to this page.

HTML

<a href="https://mcp-app-store.keenethics-labs.com/app/thebeangeek"><img src="https://mcp-app-store.keenethics-labs.com/badges/listed-on-mcp-app-store.svg" alt="Listed on MCP App Store" width="220" height="44"></a>

Markdown

[![Listed on MCP App Store](https://mcp-app-store.keenethics-labs.com/badges/listed-on-mcp-app-store.svg)](https://mcp-app-store.keenethics-labs.com/app/thebeangeek)