mcp-mealie

mcp-mealie

An MCP server for Mealie that provides curated tools for recipes, meal plans, and cookbooks with compact responses, enabling natural language management of your Mealie instance.

Category
Visit Server

README

๐Ÿฝ๏ธ mcp-mealie

An MCP server for Mealie, built for agents rather than for API coverage.

Release CI Python License: MIT Docs

Twenty-five curated tools over recipes, meal plans, cookbooks, and library cleanup, with responses trimmed hard enough that a recipe costs a few hundred tokens instead of a few thousand โ€” and sent once, not in the two copies MCP would otherwise put on the wire.

Works with any MCP client that speaks stdio: Claude Code, Claude Desktop, Cursor, Windsurf, Zed.


๐Ÿš€ Install

No clone or virtualenv needed โ€” uvx builds it straight from the tag.

{
  "command": "uvx",
  "args": [
    "--from",
    "git+https://github.com/mgummich/mcp-mealie@v0.2.1",
    "mcp-mealie"
  ],
  "env": {
    "MEALIE_URL": "https://mealie.example.com",
    "MEALIE_API_TOKEN": "your-token"
  }
}

Create the token in Mealie under Settings โ†’ API Tokens.

[!NOTE] Not on PyPI yet, so installs come from git โ€” uvx mcp-mealie on its own will not resolve. Keep the @v0.2.0 pin: without a tag you get whatever main holds the day uv resolves it.

New to this? The howto (docs/HOWTO.md) walks the whole path in order โ€” token, read-only first session, first writes, and the behaviors that surprise people once. Full docs, including the changelog, live at mgummich.github.io/mcp-mealie.

Claude Code

claude mcp add mealie \
  --env MEALIE_URL=https://mealie.example.com \
  --env MEALIE_API_TOKEN=your-token \
  -- uvx --from git+https://github.com/mgummich/mcp-mealie@v0.2.1 mcp-mealie

Claude Desktop, Cursor, Windsurf, Zed

Add the JSON block above under mcpServers in the client's config file.

From a local clone

Working on the server itself? Point the client at the checkout and every edit lands on the next restart, no push and no reinstall:

claude mcp add mealie -- uv run --directory /path/to/mcp-mealie mcp-mealie

--directory is what makes uv resolve the project from the checkout rather than from the client's working directory. Credentials can come from the repo's own .env here, so the --env flags are optional.

Updating

uvx caches the revision it first resolved, so a plain restart keeps running the old one. To move to a newer release, change the tag in the command and drop the cached build:

uv cache clean mcp-mealie    # then restart the MCP client

Restart the client after any config change; it only reads the file at startup. Released versions are listed in the changelog.

โš™๏ธ Configuration

Variable Required Default Purpose
MEALIE_URL โœ… โ€” Base URL of your Mealie instance
MEALIE_API_TOKEN โœ… โ€” Long-lived API token
MEALIE_READ_ONLY โ€” false Hide every write tool
MEALIE_VERIFY_SSL โ€” true Set false for self-signed certs (homelab only)
MEALIE_LOG_LEVEL โ€” INFO Log verbosity, to stderr

Booleans accept 1/true/yes/on and their negations. An unrecognized value is a startup error rather than a silent false.

These can also live in a .env file in the working directory (or any parent) โ€” copy .env.example to .env and fill it in. Real environment variables always take precedence over the file.

[!NOTE] Requires Mealie 2.0 or newer. The server checks at startup and refuses to run against 1.x, which has no /api/households endpoints.

๐Ÿงฐ Tools

Category Tools
๐Ÿฅ˜ Recipes search_recipes ยท get_recipe ยท suggest_recipes ยท create_recipe ยท update_recipe ยท set_recipe_image ยท upload_recipe_image ยท bulk_tag_recipes ยท delete_recipe ยท import_recipe_from_url
๐Ÿ“… Meal plans get_meal_plan ยท get_todays_meals ยท add_meal_plan_entry ยท delete_meal_plan_entry ยท random_meal_plan
๐Ÿ“š Cookbooks list_cookbooks ยท get_cookbook_recipes ยท create_cookbook ยท update_cookbook ยท delete_cookbook
๐Ÿ“Š Library reports library_stats ยท find_duplicate_recipes ยท check_recipe_links
๐Ÿ”ง Other parse_ingredients ยท manage_taxonomy

With MEALIE_READ_ONLY=true, twelve read tools remain.

Once connected, ask in plain language โ€” the agent picks the tools:

๐Ÿ’ฌ What's for dinner this week?

๐Ÿ’ฌ Import https://example.com/that-curry-recipe and tag it "Weeknight".

๐Ÿ’ฌ Plan a random week of dinners, no repeats from last week.

๐Ÿ’ฌ I have "scallion" and "spring onion" as separate foods โ€” merge them.

โœจ Things it does for you

  • Ingredients as plain text. create_recipe and update_recipe accept ["2 cups flour", "pinch of salt"] and run them through Mealie's parser. Structured objects work too; the two can be mixed. That also makes update_recipe the repair for an import that came back empty, which the import tells you about rather than leaving to be discovered later.
  • Tags by name. Mealie's API needs tag objects with a name and a slug. Pass ["Vegan"] and the server resolves or creates it, then tells you which ones were new.
  • Updates don't wipe tags. Mealie's PATCH replaces list fields wholesale. update_recipe merges tags, categories, and tools by default; pass replace_tags to overwrite. Renaming a recipe changes its slug โ€” Mealie derives one from the other โ€” so the result hands back the new one.
  • Taxonomy cleanup without a script. manage_taxonomy lists (paged, with the total), creates, renames, updates, and deletes foods, units, labels, tags, categories, and tools โ€” and merges duplicate foods or units through Mealie's own merge endpoints, so every recipe that used the loser is repointed.
  • A random week is one call. Mealie's random endpoint fills one day per request. random_meal_plan loops for you, capped at 14 days.
  • Usage rollups in one call. Mealie has no "how many recipes use this food" endpoint. library_stats("foods") sweeps the library server-side and returns the most-used foods with their counts plus every unused one โ€” the answer to "is this safe to delete" without a search per name.
  • Retag a whole shelf at once. bulk_tag_recipes(slugs, tags=["Weeknight"]) files any number of recipes through Mealie's bulk endpoints in one call, creating names that do not exist yet. It only adds; removing still goes through update_recipe with replace_tags.
  • Batch taxonomy writes. manage_taxonomy(action="update", items=[...]) runs twenty-five renames in one call, and reports per-item failures instead of stopping at the first bad id.
  • Cookbook filters without the syntax. create_cookbook(tags=["Vegan"]) builds the queryFilterString for you, matching your names to Mealie's stored casing. update_cookbook re-filters in place, so the id survives.
  • Sweeps without a call per recipe. search_recipes(fields=["slug", "tags"]) projects results the way get_recipe does, so surveying the library is one paged search rather than N follow-up reads.

๐Ÿ›ก๏ธ Safety

  • MEALIE_READ_ONLY=true prevents write tools from being registered at all.
  • delete_recipe requires the slug twice: delete_recipe(slug, confirm_slug).
  • Write requests are never retried โ€” Mealie has no idempotency key, and a retried create would duplicate the recipe.

๐ŸŽ“ Agent skill

The workflows the tool list alone doesn't teach โ€” planning a week without repeats, filing an imported recipe, writing cookbook filters, cleaning up a library rollup-first โ€” live in mealie-skill, which detects this server and drives it. It builds for Claude Code, Antigravity, Cursor, and AGENTS.md.

This repository no longer ships its own copy: two skills for one server meant two descriptions in every prompt and two places for the same guidance to drift.

๐Ÿ› ๏ธ Development

uv sync --extra dev             # creates .venv from the committed uv.lock
uv run --extra dev pre-commit install   # run the lint gates on every commit
uv run --extra dev pytest       # unit tests, fully offline
uv run --extra dev ruff check .
uv run --extra dev ruff format .
uv run --extra dev mypy         # type-checks src/

pre-commit run --all-files runs the same gates CI does.

Unit tests run entirely offline: shape.py against captured fixtures, client.py against mocked HTTP.

./scripts/integration.sh     # needs Docker: throwaway Mealie on port 19925

The integration suite spins up a real Mealie in Docker, runs tests/integration/ against it, and tears everything down. scripts/smoke.py hits a live instance of your choosing on demand.

๐Ÿ”— Related

Knuckles-Team/mealie-mcp takes the opposite approach โ€” it generates 247 tools from Mealie's OpenAPI spec, one per endpoint. Use it if you want complete API coverage. Use this one if you want a small tool list and short responses.

๐Ÿ“„ License

MIT

Recommended Servers

playwright-mcp

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.

Official
Featured
TypeScript
Magic Component Platform (MCP)

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.

Official
Featured
Local
TypeScript
Audiense Insights MCP Server

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.

Official
Featured
Local
TypeScript
VeyraX MCP

VeyraX MCP

Single MCP tool to connect all your favorite tools: Gmail, Calendar and 40 more.

Official
Featured
Local
graphlit-mcp-server

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.

Official
Featured
TypeScript
Kagi MCP Server

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.

Official
Featured
Python
E2B

E2B

Using MCP to run code via e2b.

Official
Featured
Neon Database

Neon Database

MCP server for interacting with Neon Management API and databases

Official
Featured
Exa Search

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.

Official
Featured
Qdrant Server

Qdrant Server

This repository is an example of how to create a MCP server for Qdrant, a vector search engine.

Official
Featured