Food and restaurant MCP servers are enabling AI agents to find recipes, order food delivery, make restaurant reservations, track calories, manage grocery lists, and mix cocktails — all through natural language. Instead of switching between Yelp, DoorDash, MyFitnessPal, and a recipe app, an AI agent can handle the entire food lifecycle: “Find me a chicken tikka masala recipe, check if I have the ingredients, order what’s missing from Instacart, and book a table at an Indian restaurant for Saturday in case I don’t feel like cooking.”
The landscape spans ten areas: recipes & cooking (HowToCook, Spoonacular, CookWith, TheMealDB), recipe managers (Tandoor with 3 implementations, Mealie), nutrition & food databases (OpenNutrition, USDA FoodData Central, Edamam, Nutritionix), food tracking & calories (OpenFoodFacts, calorie trackers, meal loggers), restaurants & reservations (Yelp, Resy, OpenTable, SevenRooms), restaurant operations (Toast POS), food delivery (Uber Eats, DoorDash, Swiggy, Zomato), grocery (Instacart, Kroger, Picnic, H-E-B, Whole Foods), meal kits (HelloFresh), beverages (Bar Assistant cocktails, BrewSource beer, Untappd), and multi-source food data (barcode scanning, product databases).
The headline findings: Swiggy launched Builders Club in April 2026 — opening 3 MCP servers and 18+ API tools to external developers, the first food platform to create a formal MCP developer ecosystem. Official vendor adoption remains exceptionally strong — Yelp, Instacart, Swiggy, Zomato, Spoonacular, and Edamam all have official MCP servers. HowToCook-mcp grew to 713 stars (up 25% from 569) with smart meal planning. OpenNutrition surged to 179 stars (up 47% from 122) running fully locally with 300,000+ foods. The grocery space expanded dramatically — Picnic (51 stars, 30+ tools), H-E-B (20 stars, 20+ tools), and Whole Foods all gained MCP servers. Toast POS fills the restaurant operations gap with 50+ tools. HelloFresh fills the meal kit gap. Beer gets its first MCP servers with BrewSource (BJCP styles) and Untappd.
Recipes & Cooking
HowToCook MCP
| Server | Stars | Language | License | Tools |
|---|---|---|---|---|
| worryzyy/HowToCook-mcp | 713 | TypeScript | MIT | 4 |
The most popular food MCP server — built on the viral Anduin2017/HowToCook project (“a programmer’s guide to home cooking”), which has 70,000+ stars on GitHub. Four tools: query all recipes, query by category, get specific recipe (exact + fuzzy matching), and smart meal planning that generates weekly plans based on dietary restrictions, allergies, and number of diners. Available via npm (howtocook-mcp) and now also as a DXT (Desktop Extension) for one-click Claude Desktop installation. Chinese-origin with English README.
The meal planning tool is the differentiator — it doesn’t just search recipes but creates coherent weekly meal plans accounting for nutritional balance and dietary constraints. 713 stars (up 25% from 569 in March) makes it the highest-starred food-related MCP server.
Spoonacular MCP (Official)
| Server | Language | Tools |
|---|---|---|
| ddsky/spoonacular-mcp | TypeScript | 6 |
Built by David Urbansky, the creator of the Spoonacular API itself — making this effectively an official server. Six tools: search recipes with filters, get recipe details, search ingredients by name, analyze nutrition for ingredients, find recipes by available ingredients (“what can I make with chicken and rice?"), and get random recipes. Requires a Spoonacular API key (free tier available).
A community alternative exists at zym9863/spoonacular-server in TypeScript.
CookWith MCP
| Server | Language | Tools | Notes |
|---|---|---|---|
| blaideinc/recipe-mcp | TypeScript | 4 | stdio transport, no API key (rate limited) |
| blaideinc/cookwith-mcp | TypeScript | 4 | Streamable HTTP, production endpoint |
Two implementations of the CookWith.co recipe AI — one via stdio, one via streamable HTTP (production endpoint at cookwith.co/api/mcp). Tools: generate recipes from natural language, transform/remix recipes (make it vegan, adjust calories, change serving size), dietary support, and smart nutritional adaptations. No API key required for the stdio version (rate limited). The recipe transformation capability is unique — “take this lasagna recipe and make it dairy-free with 500 calories per serving.”
Other Recipe Servers
| Server | Language | Notes |
|---|---|---|
| suraj-yadav-aiml/recipe-mcp | Python | TheMealDB API — recipe search, storage, meal plan creation |
| salindersidhu/food-mcp-server | Python | TheMealDB + USDA FoodData Central — recipes with nutritional data |
| disdjj/mcp-cook | Python | Another HowToCook dataset implementation |
| paulabaal12/kitchen-mcp | Python | Ingredient-based recipe discovery with nutrition focus |
| dmorrill/cooking-with-claude | JavaScript | Full cooking workflow — meal planning, prep, shopping, execution |
| idjohnson/recipeMakerMCP | — | Recipe creation with image generation via nanobanana |
salindersidhu/food-mcp-server stands out for combining TheMealDB recipes with USDA FoodData Central nutrition data — giving you both how to cook something and its nutritional breakdown. dmorrill/cooking-with-claude takes the most holistic approach with a full cooking workflow covering meal planning, prep, shopping, and execution.
Self-Hosted Recipe Managers
Tandoor Recipe Manager
Three separate MCP implementations for Tandoor Recipes, the popular self-hosted recipe manager:
| Server | Language | Notes |
|---|---|---|
| starbuck93/tandoor-mcp-server | TypeScript | Most comprehensive — recipes, meal plans, shopping lists CRUD, keyword/food/unit search |
| ChristopherJMiller/tandoor-mcp | Rust | Rare Rust MCP — recipes, shopping, meal planning, inventory tracking |
| mc-mario/tandoor-mcp | Python | Auto-creates missing ingredients/units/keywords — smart entity resolution |
Three implementations in three different languages (TypeScript, Rust, Python) for the same platform. mc-mario’s Python version differentiates by auto-creating missing entities (ingredients, units, keywords) when adding recipes — a small but practical feature that avoids manual setup. Note: Tandoor limits API auth to 10 requests/day, which constrains all implementations.
Mealie MCP (New)
| Server | Language | License | Tools |
|---|---|---|---|
| eds3028/mealie-mcp | Python | MIT | 7 |
A new MCP server for Mealie, another popular self-hosted recipe manager. Seven tools: recipe search with tag filtering, full recipe details, view/create meal plans, manage shopping lists, and create blank recipes. Supports both stdio (local) and SSE (remote) transport modes, with Docker support for integration into existing Mealie stacks. Requires Python 3.11+ and a running Mealie instance with API token. A welcome addition to the self-hosted recipe management space alongside Tandoor.
Nutrition & Food Databases
OpenNutrition MCP
| Server | Stars | Language | Tools |
|---|---|---|---|
| deadletterq/mcp-opennutrition | 179 | TypeScript | 3 |
The best nutrition MCP server — 300,000+ food items running fully locally with no external API calls. Three tools: food search, nutritional data lookup, and barcode scanning. Comprehensive nutritional profiles including macros, vitamins, and minerals. 179 stars (up 47% from 122 in March) — second-highest in the food MCP category. The fully-local approach means no API keys, no rate limits, no privacy concerns with dietary data, and offline capability. Data combines authoritative public sources (USDA, CNF, FRIDA, AUSNUT) for transparent, consistent nutritional data.
USDA FoodData Central Servers
Six implementations of the USDA FoodData Central API — the US government’s comprehensive food composition database:
| Server | Language | Notes |
|---|---|---|
| i-am-neon/nutrition-mcp-server-python | Python | USDA lookup + automated meal planner with diet-specific prompts |
| asachs01/nutrition-mcp | JavaScript | 24-hour API cache, local meal log, nutrition goal targets |
| zen-apps/mcp-nutrition-tools | Python | Dual architecture (MCP + HTTP API), Docker/Cloud Run ready |
| jlfwong/food-data-central-mcp-server | TypeScript | Direct USDA API integration |
| HemantaPatil/mcp_postgres | Python | USDA data in local PostgreSQL, MCP Inspector web interface |
| MCP-Forge/nutritionix-mcp-server | Python | Nutritionix API — natural language exercise calorie estimates |
This is the most duplicated subcategory — 6 servers wrapping the same government API. i-am-neon’s version adds an automated meal planner. asachs01’s adds meal logging with goals. zen-apps’ has a dual MCP + HTTP architecture for cloud deployment. For most users, mcp-opennutrition (above) is a better choice since it runs locally with a larger database.
Official Food Data APIs
| Server | Notes |
|---|---|
| Edamam MCP | Official — food search by name or barcode (UPC/EAN/PLU), nutrient analysis, food image analysis via AI, REST + JSON-RPC + Inspector UI |
| GuptaPurujit/nutritionix_mcp | Nutritionix API v2 — natural language meal logging, calorie/macro breakdowns |
| fcoury/fatsecret-mcp | FatSecret — OAuth 1.0a, food diary logging |
Edamam’s official MCP is notable for food image analysis — photograph your plate and get nutritional estimates. This is a genuinely useful AI-native feature that goes beyond simple database lookups.
Food Tracking & Calories
Calorie & Meal Trackers
| Server | Language | Notes |
|---|---|---|
| nagarjun226/food-tracker-mcp | Python | OpenFoodFacts — barcode scanning, dietary restriction management, consumption logging |
| neonwatty/food-tracker-mcp | Python | USDA — food logging, daily totals vs goals, local SQLite |
| thitiph0n/calorie-tracker-mcp-server | TypeScript | Cloudflare Workers + D1 — BMR/TDEE via Mifflin-St Jeor, body composition tracking, OAuth 2.1 with PKCE |
| VeriTeknik/daily-calorie-tracker | TypeScript | Built-in database of 100+ common foods, natural language entry, weekly reports |
| ajaykallepalli/MCP_Food_Search | Python | Grocery price search + nutritional info + cross-store price comparison |
| evangstav/personal-mcp | Python | Multi-purpose server — nutrition module logs meals with hunger/satisfaction levels |
thitiph0n/calorie-tracker-mcp-server is the most production-ready — built on Cloudflare Workers with D1 database, OAuth 2.1 with PKCE authentication, BMR/TDEE calculations using the Mifflin-St Jeor equation, and historical body composition tracking. ajaykallepalli/MCP_Food_Search stands out for combining nutrition data with grocery pricing and cross-store price comparison.
OpenFoodFacts Servers
Four implementations of the OpenFoodFacts database — the Wikipedia of food products:
| Server | Language | Notes |
|---|---|---|
| JagjeevanAK/OpenFoodFacts-MCP | Python | 7 tools — AI-powered analysis, product comparison, recipe suggestions |
| noot-app/openfoodfacts-mcp-server | TypeScript | Parquet database with DuckDB — local, automatic dataset updates, file locking |
| caleb-conner/open-food-facts-mcp | Python | Nutri-Score, NOVA, Eco-Score analysis — Docker-ready |
noot-app’s TypeScript version is architecturally interesting — it uses Parquet files with DuckDB for local querying rather than hitting the OpenFoodFacts API, with automatic dataset updates and file locking for concurrent safety. caleb-conner’s version includes Nutri-Score, NOVA food processing classification, and Eco-Score environmental impact ratings.
Restaurants & Reservations
Yelp MCP (Official)
| Server | Language | Tools |
|---|---|---|
| Yelp/yelp-mcp | Python | 1 (agent-to-agent) |
Official Yelp product — a single yelp_agent tool with a unique agent-to-agent communication design. Instead of exposing multiple search/filter tools, you send natural language queries and the Yelp agent handles parsing, search, comparison, and itinerary planning. Supports follow-up via chat_id for multi-turn conversations. Uses the Yelp Fusion AI API.
The agent-to-agent design is architecturally significant — rather than an MCP server being a thin API wrapper, Yelp’s server is itself an AI agent that understands context and maintains conversation state. This points toward a future where MCP servers are increasingly agentic.
Restaurant Reservation Servers
| Server | Language | Notes |
|---|---|---|
| jrklein343-svg/restaurant-mcp | TypeScript | Unified Resy + OpenTable — parallel search, reservation sniper, DPAPI-encrypted credentials |
| samwang0723/mcp-booking | Go | Google Maps integration — AI recommendations, mood-based filtering |
| OpenTable MCP (Apify) | — | Scraper-based OpenTable search and availability |
| SevenRooms MCP | — | Enterprise restaurant reservation management |
| Resy MCP (LobeHub) | — | Search, availability, reservations, waitlists |
jrklein343-svg/restaurant-mcp is the standout — it searches both Resy and OpenTable simultaneously, comparing results in parallel. The reservation sniper feature continuously monitors for table availability and books instantly when a slot opens. DPAPI-encrypted credential storage shows security awareness — the server never handles payment information. samwang0723/mcp-booking takes a mood-based approach: describe the occasion (“anniversary dinner” or “casual Friday with friends”) and it recommends restaurants from Google Maps.
Restaurant Operations (New)
Toast POS MCP
| Server | Stars | Language | License | Tools |
|---|---|---|---|---|
| BusyBee3333/toast-mcp-2026-complete | 2 | TypeScript | MIT | 50+ |
The first restaurant POS MCP server — a community-built integration with Toast, the leading restaurant management platform. 50+ tools across 10 categories: order lifecycle management (12 tools), menu and inventory operations (11 tools), payment and cash management (8 tools), sales reporting and analytics (6 tools), staff and labor analytics (6 tools), and customer relationship management (4 tools). Implements OAuth 2.0 authentication with the Toast API and provides both stdio mode (MCP integration) and HTTP mode (web UI).
This fills a significant gap from the original review — restaurant operations and POS integration. While community-built rather than official, the breadth of 50+ tools covering the full restaurant operations lifecycle (orders, menus, payments, staff, analytics, CRM) makes it a serious entry. No Square POS MCP server has appeared yet.
Food Delivery
Uber Eats MCP
| Server | Stars | Language | Tools |
|---|---|---|---|
| ericzakariasson/uber-eats-mcp-server | 221 | Python | — |
The most popular food delivery MCP server at 221 stars (up slightly from 217). A proof-of-concept using browser automation for restaurant search, menu exploration, and food ordering. While not production-grade, it demonstrates the potential for AI-driven food ordering workflows.
DoorDash Servers
| Server | Language | Notes |
|---|---|---|
| JordanDalton/DoorDash-MCP-Server | TypeScript | DoorDash API integration, requires DOORDASH_API_KEY |
| amannm/doordash-mcp | TypeScript | Minimal MCP-to-DoorDash-SDK — delivery quotes, creation, status, cancellation |
| Doriandarko/webmcp-starter | — | Educational demo — DoorDash-style delivery workflow with 9 AI tools |
amannm/doordash-mcp is the most focused — a minimal transformation layer between MCP and the DoorDash SDK with delivery quotes, creation, status checking, and cancellation. Unlike Uber Eats, DoorDash has a developer API, so these servers use official SDK endpoints rather than browser automation.
Official Food Delivery Platforms
| Server | Notes |
|---|---|
| Swiggy MCP | Official — three services: swiggy-food (restaurant ordering), swiggy-instamart (grocery/quick-commerce, 40,000+ SKUs), swiggy-dineout (restaurant discovery). Launched January 2026. First quick-commerce platform globally to adopt MCP. |
| Zomato MCP | Official — food ordering, order tracking, QR code payment. Production endpoint at mcp-server.zomato.com. Integrates with ChatGPT and Claude. |
| GrubHub Scraper MCP | Apify-based scraper, not official |
Swiggy and Zomato — India’s two largest food delivery platforms — both have official MCP servers, making India the most MCP-advanced food delivery market. Swiggy’s Instamart integration is particularly notable as the first quick-commerce (instant grocery delivery) MCP globally, covering 40,000+ SKUs via natural language. Zomato’s QR code payment feature shows how MCP can bridge digital ordering with physical restaurant experiences.
April 2026 update: Swiggy Builders Club — Swiggy announced Builders Club, a developer programme opening its AI commerce infrastructure to external developers, startups, and enterprises. Built on AWS (Amazon Bedrock and AgentCore), the programme gives approved builders access to 3 MCP servers and 18+ API tools spanning Food, Instamart, and Dineout. Builders can create AI agents, copilots, and integrations that take real-world actions — ordering food, shopping groceries, booking dining. Includes a “Skills framework” of reusable capabilities for agentic commerce. This makes Swiggy the first food delivery platform to create a formal developer ecosystem around MCP, with live API access, generous rate limits, and direct engineering support. A significant milestone for the entire food MCP space.
Multi-Platform Delivery
| Server | Notes |
|---|---|
| MCP A2A AP2 “I’m Hungry” (Glama) | Multi-delivery service discovery across DoorDash, UberEats, and GrubHub with Stripe payment processing and cryptographically signed user authorization |
Grocery Shopping
Instacart MCP (Official)
| Server | Notes |
|---|---|
| Instacart MCP | Official — create recipes (with ingredients, instructions, measurements), create shopping lists, discover retailers by postal code. Shareable URLs for recipes and lists. |
| sjufan84/instacart-cli | Go CLI + library for Instacart recipe and shopping list creation |
Instacart’s official MCP creates a compelling workflow: describe a meal, get a recipe page with ingredients and instructions, then convert it to a shopping list with one-click ordering. The shareable URL feature means an AI agent can create a recipe, generate a link, and hand it to you for review before placing an order.
Picnic Supermarket MCP (New)
| Server | Stars | Language | Tools |
|---|---|---|---|
| ivo-toby/mcp-picnic | 51 | TypeScript | 30+ |
An impressive MCP server for Picnic, the European online supermarket delivery service (Netherlands, Germany, France). 30+ tools covering product search, cart management, order placement, meal planning with automatic shopping list generation, delivery tracking and optimization, budget management with spending analysis, and transaction history. Supports session persistence with 2FA, works with Claude Desktop and other MCP clients. Available via npm (mcp-picnic). 51 stars makes it the most popular grocery-specific MCP server.
An alternative exists at YussufSassi/picnic-mcp.
H-E-B Grocery MCP (New)
| Server | Stars | Language | License | Tools |
|---|---|---|---|---|
| mgwalkerjr95/texas-grocery-mcp | 20 | Python | MIT | 20+ |
A comprehensive MCP server for H-E-B, the beloved Texas grocery chain. 20+ tools across five categories: store location (3 tools), product search with real-time pricing (3 tools), cart management with confirmation requirements (5 tools), digital coupon discovery and clipping (4 tools), and session management (4 tools). The coupon clipping feature is unique — no other grocery MCP server offers automated deal discovery. Uses unofficial web APIs and browser automation. Note: unofficial integration.
Whole Foods MCP (New)
| Server | Language | Tools |
|---|---|---|
| benjiebob/whole-foods-mcp | Python | 8 |
The first Whole Foods MCP server — enables automated grocery ordering from Amazon Whole Foods via browser automation. Eight tools: login, session saving, product search, product details, add to cart, view cart, remove from cart, and clear cart. Supports weight-based items (produce). Requires Python 3.11+ and an Amazon account with Whole Foods delivery enabled. Fills a significant gap from the original review.
Dutch Supermarket Price Comparison (New)
| Server | Stars | Language | License |
|---|---|---|---|
| Samvox1/nl-supermarkt-mcp | 5 | Python | MIT |
Searches and compares prices across 12+ Dutch supermarket chains (Albert Heijn, Jumbo, Lidl, Aldi, Plus, DekaMarkt, Dirk, and more) plus 7 drugstore chains (Kruidvat, Etos, etc.). Features include shopping list optimization to find the cheapest combinations, weekly promotion tracking, price history alerts, budget management, and weekly menu planning with automatic grocery generation. A creative multi-vendor approach.
Other Grocery Servers
| Server | Stars | Language | License | Notes |
|---|---|---|---|---|
| CupOfOwls/kroger-mcp | 4 | Python | MIT | Product search, store finder, cart management, shopping path optimization |
| jessalva7/grocery-mcp-server | — | Python | — | General grocery item and shopping list management |
| o-b-one/groceries-mcp | 3 | Python | MIT | Multi-vendor grocery API (Rami Levy, Keshet), cart automation |
CupOfOwls/kroger-mcp has a unique shopping path optimization tool — it finds the optimal route through a Kroger store based on your shopping list. While a niche feature, it demonstrates how MCP servers can provide utility beyond simple API wrappers.
Meal Kits (New)
HelloFresh MCP
| Server | Notes |
|---|---|
| striderlabs/mcp-hellofresh | Weekly menu browsing, recipe details (ingredients, instructions, nutrition, allergens), delivery schedule management, subscription plan updates, dietary preference editing |
The first meal kit MCP server — enables AI agents to manage HelloFresh accounts end-to-end. Browse current and future weekly menus, access detailed recipe information including allergens and nutrition, view and modify delivery schedules, and update subscription plans and dietary preferences. Install via claude mcp add hellofresh -- npx -y @striderlabs/mcp-hellofresh. Requires HelloFresh account credentials.
An alternative by markswendsen-code uses Playwright browser automation for the same functionality. This fills a gap called out in the original review — no Blue Apron or Factor MCP servers yet, but HelloFresh as the market leader is the most impactful entry.
Beverages
Bar Assistant MCP
| Server | Language | Tools |
|---|---|---|
| zhdenny/bar-assistant-mcp-server | Python | 7+ |
Smart cocktail search with natural language + similarity matching, complete recipe retrieval, ingredient analysis, batch recipe retrieval, inventory-based recommendations, and advanced filtering by ABV, glassware, and preparation method. The similarity engine is clever — search for “something like a Negroni but less bitter” and it finds cocktails with similar profiles. 70%+ cache hit rate for performance. Integrates with Bar Assistant instances.
BrewSource Beer MCP (New)
| Server | Language | Tools |
|---|---|---|
| CharlRitter/brewsource-mcp | Go | 3 |
The first beer-focused MCP server — provides AI assistants with brewing knowledge and beer discovery. Three tools: bjcp_lookup for Beer Judge Certification Program style specifications (the standard reference for beer styles), search_beers for commercial beer search by name/style/brewery/location, and find_breweries for brewery discovery with geographic filtering. Also exposes 5 MCP resources including complete BJCP style guidelines (beer, mead, cider), commercial beer catalog, and brewery directory. Currently in Phase 1 MVP. Uses PostgreSQL for data and Redis for caching. Future phases will add ingredient databases, personal recipe analytics, and Brewfather API integration.
Untappd MCP (New)
| Server | Language | License | Tools |
|---|---|---|---|
| jtucker/mcp-untappd-server | JavaScript | GPL-3.0 | 3 |
MCP server for the Untappd beer social network — three tools: beer search, detailed beer info, and user checkins. Important caveat: Untappd is no longer accepting registrations for new API keys, limiting this to developers with existing access. The get_user_checkins tool is noted as non-functional.
The beverage space has significantly improved since the original review — from a single cocktail server to cocktails, beer styles, and beer social, though wine remains absent.
What’s Missing
Despite 70+ servers (up from 55+), several gaps have been filled since March while others remain:
Gaps filled since last review:
No restaurant POS/operations→ Toast POS MCP now has 50+ tools (community-built)No Whole Foods→ benjiebob/whole-foods-mcp provides browser-automated orderingNo meal kit services→ HelloFresh MCP for account management and menu browsingNo beer databases→ BrewSource MCP (BJCP styles, brewery/beer search) and Untappd MCP (limited by API key closure)
Remaining gaps:
- No official DoorDash or GrubHub MCP servers — DoorDash has a developer API but no official MCP server; GrubHub only has third-party scrapers
- No Walmart Grocery, Amazon Fresh, or Target — major US grocery platforms still absent (Whole Foods partially fills this via Amazon)
- No dietary condition management — no celiac disease, diabetes, IBS, or allergy-aware meal planning with medical-grade data
- No Square, Clover, or Lightspeed POS — Toast is covered but the broader restaurant POS market is not
- No food safety or allergen databases — no FDA food recalls, no comprehensive allergen cross-reference databases
- No food photography or plating — no food styling, plating suggestions, or photo enhancement tools
- No wine databases — no Vivino, CellarTracker, or wine-focused servers (beer is now covered but wine is not)
- No food waste management — no servers for tracking expiration dates, suggesting recipes for ingredients about to expire
- No restaurant review aggregation beyond Yelp — no Google Maps reviews, TripAdvisor, or The Infatuation
- No farm-to-table or sourcing — no farmers market directories, CSA boxes, or local food sourcing
- No Blue Apron, Factor, or other meal kits — HelloFresh leads but the broader meal kit market is uncovered
The Bottom Line
Food & Restaurant MCP servers earn 4.0 out of 5. The April 2026 headline is Swiggy launching Builders Club — opening 3 MCP servers and 18+ API tools to external developers, making it the first food platform to create a formal MCP developer ecosystem. This represents a maturation from “we have an MCP server” to “build your business on our MCP infrastructure.”
Official vendor participation remains the category’s strength — Yelp, Instacart, Swiggy, Zomato, Spoonacular, and Edamam all have official MCP servers. The recipe space is led by HowToCook-mcp (713 stars, up 25%) with Mealie adding self-hosted recipe management. OpenNutrition surged 47% to 179 stars as the best local-first nutrition database.
The biggest improvement since March is gap-filling: Toast POS (50+ tools) fills restaurant operations, Whole Foods fills a major US grocery gap, Picnic (51 stars, 30+ tools) and H-E-B (20 stars, 20+ tools) expand international and regional grocery coverage, HelloFresh fills meal kits, and BrewSource fills beer. The grocery space in particular transformed from 2 servers (Instacart, Kroger) to 7+ covering the US (Whole Foods, Kroger, H-E-B), Europe (Picnic NL/DE/FR, Dutch multi-supermarket), and Israel (Rami Levy).
The weaknesses are the continued duplication in nutrition tracking, the absence of official DoorDash/GrubHub servers, and the remaining gaps in wine, dietary condition management, and food safety databases. The Western food delivery market still lags India’s production-grade official implementations.
This review was refreshed on 2026-04-26 using Claude Opus 4.6 (Anthropic).