willhaben MCP server: let your AI agent search the Austrian marketplace

#What your agent gets

Wield exposes willhaben through wield-mcp, a local MCP server. Your agent finds the recipe with find_recipe and runs it with run. A recipe is a small public file that describes the requests willhaben's own search page makes.

The search returns the total number of matching ads, the current page and up to 30 items per page. Each item has a title, price in euros, a free flag, price text, location, postcode, state, publish time, a private flag, the listing link, an image link and a description.

You can filter by keyword, minimum and maximum price, and Bundesland (state). You can sort by relevance, newest first, cheapest first or most expensive first. For the exact params and example prompts, read the willhaben how-to page.

#How a request flows

1. You ask your agent: "Find a used bicycle in Wien between 50 and 300 euros, cheapest first."
2. The agent calls find_recipe, then run with willhaben.at/search-marketplace.
3. wield-mcp hands the run to the Wield extension through a local bridge. The bridge uses Chrome native messaging, so there is no network port.
4. The extension sends one request from a willhaben.at tab. The request carries your own browser session.
5. The results return to your agent. Nothing passes through wield.run.

The agent is told to treat results as website content. It does not follow instructions it finds inside them.

#What leaves your computer

One search request goes to www.willhaben.at per run, sent from your own browser tab. Wield logs the method, host, path, status and timing on your computer. It never logs request bodies or query values. Telemetry is off by default. Wield never reads, copies or sends your cookies or passwords.

#willhaben terms, in plain words

willhaben's terms prohibit automated reading of ads. The recipe runs one search page at a time, and you start it. It does not contact sellers. Contacting a seller stays on willhaben.at. Read willhaben's terms yourself. Wield's recipes note known restrictions, and this one is listed above.

#Limits and what is not working yet

• Bot checks stop the run. If willhaben shows a check, such as a CAPTCHA or an "are you human" page, the run stops with site_challenge. You solve the check in your own tab. Wield does not solve bot protection.
• Recipe status. The recipe is verified by a live check on 7 November 2026. Recipes are health-checked daily. If a check fails, the recipe is marked degraded, and wield-mcp falls back to the last good version.
• Limits per run. A run stops after 12 steps, 40 requests, 5 MB per response or 60 seconds.
• Scope. This recipe covers willhaben's marketplace search. It does not message sellers or place offers.

#FAQ

#Is there an official willhaben MCP server?

Not that we know of as of 7 November 2026. Wield is not part of willhaben. It is a local MCP server that runs a public recipe for willhaben's marketplace search.

#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. Other MCP clients work through the standard stdio config. See AI apps that work with Wield.

#Does it need a login?

No. The search recipe has no login step. You can search willhaben without an account. Wield never reads or sends your cookies or passwords.

#How fast is it compared with a browser agent?

In our benchmark, as of 7 November 2026 (preliminary, still running), the Wield recipe passed 9 of 9 runs with a median of 582 ms and no model cost. A Browser Use agent driving Claude Sonnet 5.5 passed 4 of 4 runs with a median of 162.0 s and $0.47 per run. The task check was at least five listings with a title and a price. The full method is on the benchmark page.