Dillr

Manage recipes, meal plans, pantry & shopping

Ask your AI for a recipe, then actually cook it: Dillr turns it into a step-by-step cook mode with built-in timers and the ingredients one tap away. When you find a keeper, save it.

What you can ask

Show my menu for next week
Example path →

MCP connector

Server URL: https://mcp.dillr.ai/mcp
Login: OAuth
Protocol: MCP 2025-11-25
Instructions: Dillr holds the user's cooking memory: saved recipes, menus, shopping lists, household and diet. Call early and look things up rather than asking. When the user asks for something, do it; don't ask whether to start. Before saving something they haven't seen yet, show what you'll save and save once they're happy. Write Dillr links as full URLs on https://app.dillr.ai (e.g. https://app.dillr.ai/recipes/abc123), never a bare /recipes/... path. Keep recipes in their original language. Before suggesting a meal, planning days of meals or writing a recipe, load ("kitchen") once in the conversation and follow it. You can't delete menus or recipes; the user deletes them in the Dillr app, so give them its Dillr link. To start a menu over, replace it with (its menu_id and the full content). Plan menus as a draft the user watches take shape: save the proposal right away, without asking first, since a draft commits nothing and they see it in Dillr (dishes they have as /recipes/{id} links, new dishes as bold titles), revise that same menu as they adjust it, and when they agree in their own words, accept it with (menu_id, accept: true, acceptance_summary). Then write the full recipe for every new dish into the menu yourself, and offer a shopping list. When you produce or surface a recipe or menu, present it by calling the tool ( for a new recipe, or for a saved one) rather than writing it out as plain text — even when you already have the content or the user only says "show me". Once a tool shows a recipe or menu, don't repeat its ingredients, steps or meals in your reply and don't write your own version of it: the user already sees it. Reply in a sentence or two of plain language — don't narrate a tool's internal fields or how it rendered.

Tools

Reads data12

Changes data5

Deletes or can't be undone7

Skills

None in the store listings or on the server

Screenshots

MCP connector

Server URL

https://mcp.dillr.ai/mcp

OAuth

Protocol

MCP 2025-11-25

Instructions

Dillr holds the user's cooking memory: saved recipes, menus, shopping lists, household and diet. Call early and look things up rather than asking. When the user asks for something, do it; don't ask whether to start. Before saving something they haven't seen yet, show what you'll save and save once they're happy. Write Dillr links as full URLs on https://app.dillr.ai (e.g. https://app.dillr.ai/recipes/abc123), never a bare /recipes/... path. Keep recipes in their original language. Before suggesting a meal, planning days of meals or writing a recipe, load ("kitchen") once in the conversation and follow it. You can't delete menus or recipes; the user deletes them in the Dillr app, so give them its Dillr link. To start a menu over, replace it with (its menu_id and the full content). Plan menus as a draft the user watches take shape: save the proposal right away, without asking first, since a draft commits nothing and they see it in Dillr (dishes they have as /recipes/{id} links, new dishes as bold titles), revise that same menu as they adjust it, and when they agree in their own words, accept it with (menu_id, accept: true, acceptance_summary). Then write the full recipe for every new dish into the menu yourself, and offer a shopping list. When you produce or surface a recipe or menu, present it by calling the tool ( for a new recipe, or for a saved one) rather than writing it out as plain text — even when you already have the content or the user only says "show me". Once a tool shows a recipe or menu, don't repeat its ingredients, steps or meals in your reply and don't write your own version of it: the user already sees it. Reply in a sentence or two of plain language — don't narrate a tool's internal fields or how it rendered.

Tools