Sign inTry it

WG-Gesucht MCP server: let your AI agent search WG rooms and flats

Wield is an MCP server that gives your AI agent one WG-Gesucht tool, wg-gesucht.de/search-rooms. The tool searches WG rooms, one-room flats, flats and houses in a German city, with a maximum rent and a minimum size. Each offer shows the rent, size, flatmates, district, availability and link. It runs in your own browser, and the recipe needs no login.

Updated

Live recipewg-gesucht.de/search-roomsv1

WG-Gesucht Find a WG room or flat on WG-Gesucht

Read-only, no loginRuns in your browser only

Parameters (6)
categorystring "wg_room" | "one_room_flat" | "flat" | "house"
citystringrequiredCity, e.g. 'Berlin', 'München', 'Frankfurt am Main', 'Wien'. The first match of WG-Gesucht's city search is used
include_partner_offersbooleanKeep offers from commercial partners (Sofortbuchung) and paid placements, marked
max_rentintegerEUR per month, total rent
min_sizeintegerm²
pageinteger
Use it with your agent

What your agent gets

RecipeWhat it doesLoginStatus (7 Nov 2026)
wg-gesucht.de/search-roomsWG rooms, one-room flats, flats and houses in a city, with max rent, min size and availability, newest firstNoOK, 6 of 6 checks, 2 requests

The tool takes city as its only required input. You can add category (wg_room by default, one_room_flat, flat or house), max_rent in EUR per month for the total rent, min_size in m², page and include_partner_offers. The city is looked up with WG-Gesucht's own city search. The result shows the city it matched, for example München (Bayern).

In our check on 7 November 2026, a Munich search with a maximum rent of 700 EUR and a minimum size of 15 m² passed 6 of 6 checks in 2 requests and 0.7 seconds. A run has no model in the loop, so it uses no LLM tokens.

How a request flows

  1. You ask your agent: "Find WG rooms in München up to 700 EUR a month, at least 15 m², newest first."
  2. The agent calls find_recipe, then run with wg-gesucht.de/search-rooms and the parameters.
  3. wield-mcp passes the run through the local bridge to the Wield extension.
  4. The extension sends the requests from your browser, so the search runs the way it would if you searched the site yourself.
  5. The offers go back to your agent. Nothing passes through wield.run.

What leaves your computer

The only host is www.wg-gesucht.de. A search makes two requests: one for the city lookup and one for the list page. The recipe needs no login and sends no messages. Wield logs each request on your computer only: method, host, path, status and timing. It never logs bodies or query values. Telemetry is off by default. Wield never reads, copies or sends your cookies or passwords.

WG-Gesucht's terms, in plain words

We have not reviewed WG-Gesucht's terms for automated access yet, so this is not legal advice. One run is one person's search, with two requests, the same ones the city box and the list page make. Contacting landlords stays on wg-gesucht.de. Check the site's terms yourself before you automate anything more.

Limits and what is not working yet

  • Read only. The recipe sends no messages to landlords and needs no login.
  • Partner offers are hidden. Offers from commercial partners and paid placements are dropped unless you set include_partner_offers. Kept offers are marked partner_offer.
  • Advertisers' names are left out of the result.
  • Cities are matched, not confirmed. The first match wins, so check city in the result.
  • Run limits apply. A run has at most 12 steps, 40 requests, 5 MB per response and 60 seconds.

FAQ

Does it need a WG-Gesucht login?

No. The search works without an account.

Can the tool message a landlord?

No. The recipe reads offers only. Contacting landlords stays on wg-gesucht.de.

Which AI apps work with it?

Any MCP client. npx -y wield-mcp init sets up Claude Desktop, Claude Code, Cursor and Codex. Claude Desktop also has a one-click bundle on wield.run/install.

Are partner offers included?

No, by default. Set include_partner_offers to true to keep them. They stay marked as partner offers.

How does it compare with a browser agent?

We have not benchmarked this recipe yet, so this page makes no speed or success comparison.