Abfall
FreeNot checkedMCP server for the waste collection calendars of German cities and districts - 995 waste authorities.
About
MCP server for the waste collection calendars of German cities and districts - 995 waste authorities.
README
English · Deutsch
CI License: MIT Python 3.11+ Providers: 995
MCP server for the waste collection calendars of German cities and districts. Looks up collection dates — residual waste, organic waste, paper, packaging, bulky waste, hazardous waste collection points — for a given address.
Website and interactive setup guide: abfall-mcp.alpaycelik.dev.
address ──▶ Nominatim ──▶ municipality + district ──▶ provider search (995)
│
collection dates ◀── portal ◀── resolve arguments
Why this is not trivial
Waste management in Germany is organised municipally. There is no nationwide API — roughly 400 public waste authorities each run their own portal, backed by a handful of software platforms (Abfall.IO/AbfallPlus, AbfallNavi, Jumomind, AWIDO, C-Trace, Müllmax, plus many plain ICS exports).
This server uses mampfes/hacs_waste_collection_schedule (MIT) as its data source — an actively maintained project with over 950 source modules. Its inner package is independent of Home Assistant and is wired in here as a Git submodule.
The actual work of this project is turning an address into the responsible waste authority and its internal parameters. In Home Assistant a human clicks that together once; an MCP server has to do it on its own.
Installation
The server runs from a repository checkout, not as an installed package: it
needs the data source under vendor/ and the registry under data/, both
resolved relative to the project root. That is also why it is not on PyPI — a
wheel would install cleanly and still not work.
git clone --recurse-submodules https://github.com/AlpayC/abfall-mcp-server.git
cd abfall-mcp-server
uv sync
uv run pytest
data/providers.json is checked in, so the server starts right away. After a
submodule update, rebuild it:
uv run python scripts/build_registry.py
Wiring it up
Public remote via Streamable HTTP:
{
"mcpServers": {
"abfall": {
"url": "https://abfall-mcp.alpaycelik.dev/mcp"
}
}
}
The service health endpoint is https://abfall-mcp.alpaycelik.dev/health.
Locally over stdio — in claude_desktop_config.json or .mcp.json:
{
"mcpServers": {
"abfall": {
"command": "uv",
"args": ["--directory", "/path/to/abfall-mcp-server", "run", "abfall-mcp-server"]
}
}
}
As an HTTP service:
uv run abfall-mcp-server --http --host 127.0.0.1 --port 8000
Keep in mind that every HTTP request puts load on the authorities' portals and
on Nominatim. For anything beyond personal use, put a cache in front and your
own Nominatim instance behind it (ABFALL_MCP_NOMINATIM_URL).
Tools
Tool names and responses are German, because the domain and the users are.
| Tool | Purpose |
|---|---|
abfuhrtermine |
Address in, collection dates out. The usual entry point. |
finde_traeger |
Search authorities by place or company name, no geocoding. |
traeger_details |
Which arguments does an authority expect? |
abfuhrtermine_fuer_traeger |
Targeted query, e.g. to answer a follow-up question. |
abdeckung |
How many authorities and data sources are covered. |
Plus the resource abfall://traeger with the full list of authorities.
Asking beats guessing
When a value cannot be determined unambiguously, the server returns
status: "rueckfrage" along with the concrete list of options instead of
guessing. That is deliberate: a wrongly guessed town cheerfully returns the
neighbouring municipality's calendar — a wrong answer that looks like a right
one.
Coverage, measured
scripts/build_registry.py collects 995 waste authorities from 150 source
modules.
scripts/smoke.py queries real addresses against the real portals. A run over
21 addresses across Germany (as of Aug 2026):
| Outcome | Share |
|---|---|
| Collection dates returned directly | 48 % |
| Follow-up question for a missing detail | 14 % |
| No authority returned dates | 38 % |
That is the measured number, not an estimate — and the reason to state it here: "covers every German city" is true of the authority list, not of fully automatic resolution from a bare address.
Authorities with their own ID lookup
Some portals require internal identifiers that cannot be derived from an
address. For these, the portal's address dialog is reimplemented in
lookup.py:
| Authority | Identifier | Scope |
|---|---|---|
| Abfall.IO / AbfallPlus | f_id_kommune, f_id_strasse, … |
41 authorities |
| Stadtreinigung Hamburg | hnId |
Hamburg |
| Berliner Stadtreinigungsbetriebe | schedule_id |
Berlin |
The upstream wizard for Hamburg is stale by now — the portal moved its form to a JavaScript component, whose address endpoint is read out of the page here.
Where the remaining cases fail
- Further ID arguments with no lookup path:
standortin Dresden,idHouseNumberin Leipzig,streetnrin Stuttgart. One more resolver each, following the same pattern as the three above. - Portals using different spellings that ship no list of suggestions (Erfurt, Kiel).
- Outages, rate limiting and malformed responses on the authorities' side (Saarbrücken returns HTML instead of ICS).
Layout
| File | Responsibility |
|---|---|
wcs.py |
Bridge to the vendored library; registers the package deliberately instead of putting it on sys.path (its parent directory holds a calendar.py that shadows the stdlib). |
registry.py |
Authority list and location search with German stemming. |
geo.py |
Nominatim, address variants, plausibility checks. |
resolve.py |
Address → authority → resolved arguments → collection dates. |
lookup.py |
Address dialogs for authorities with internal IDs (Abfall.IO, Hamburg, BSR). |
server.py |
MCP tools, stdio and HTTP. |
The registry is not built at runtime: data/providers.json is produced by a
script so that starting the server does not import 150 modules. Rebuild it
after a submodule update.
Data sources and usage
Collection dates come from the portals of the respective waste authorities,
address resolution from Nominatim
(OpenStreetMap). Nominatim has a usage policy — at most one request per second;
the server honours it and caches results in ~/.cache/abfall-mcp-server/.
For dates something depends on (bulky waste, hazardous waste collection), it is worth checking the portal address that every response carries.
Contributing
The most useful contribution is a report that some authority does not work — there is an issue template with the right questions. How to add an authority resolver is in CONTRIBUTING.md; security issues belong in a private report, see SECURITY.md. Changes are listed in the CHANGELOG. Agents working on this repository should read AGENTS.md.
One principle runs through the whole project and applies to contributions too: when in doubt, ask — do not guess. A wrongly guessed town cheerfully returns the neighbouring municipality's calendar — a wrong answer that looks like a right one.
License
MIT, see LICENSE. The submodule
vendor/hacs_waste_collection_schedule is under its own MIT license,
Copyright (c) 2020 Steffen Zimmermann — this repository only references it, it
does not ship the code. Third-party attributions are collected in
NOTICE.
Install Abfall in Claude Desktop, Claude Code & Cursor
unyly install abfallInstalls into Claude Desktop, Claude Code, Cursor & VS Code — handles npx, uvx and build-from-source repos for you.
First time? Get the CLI: curl -fsSL https://unyly.org/install | sh
Or configure manually
Run in your terminal:
claude mcp add abfall -- uvx --from git+https://github.com/AlpayC/abfall-mcp-server abfall-mcp-serverStep-by-step: how to install Abfall
FAQ
Is Abfall MCP free?
Yes, Abfall MCP is free — one-click install via Unyly at no cost.
Does Abfall need an API key?
No, Abfall runs without API keys or environment variables.
Is Abfall hosted or self-hosted?
Self-hosted: the server runs locally on your machine via the install command above.
How do I install Abfall in Claude Desktop, Claude Code or Cursor?
Open Abfall on unyly.org, pick your client tab (Claude Desktop, Claude Code, Cursor) and press Install — the config is generated automatically, no JSON editing.
Related MCPs
GitHub
PRs, issues, code search, CI status
by GitHubFilesystem
Secure file operations with configurable access controls.
Memory
Knowledge graph-based persistent memory system.
Template MCP Server
A CLI tool to create a new Model Context Protocol server project with TypeScript support, dual transport options, and an extensible structure
by mcpdotdirectAmap Maps Mcp Server
MCP server for using the AMap Maps API
by duxiaohuiSupabase
Database, auth and storage
by SupabaseEverything
Reference / test server with prompts, resources, and tools.
Git
Tools to read, search, and manipulate Git repositories.
Sequential Thinking
Dynamic and reflective problem-solving through thought sequences.
Time
Time and timezone conversion capabilities.
Compare Abfall with
Not sure what to pick?
Find your stack in 60 seconds
Author?
Embed badge for your README
Browse similar
All development MCPs
