Lieferando MCP server: let your AI agent search restaurants that deliver to an address
Wield is an MCP server that lets your AI agent search Lieferando restaurants that deliver to an address. The search runs in your own browser with your own session, so requests come from your machine, not a cloud server. The search needs no login. A second tool, order status, exists but is untested because it needs a logged-in recording.
Lieferando Search restaurants that deliver to an address
Read-only, no loginNo reports yet
Parameters (8)
| Metric | Wield recipe | Browser Use + Claude Sonnet 5.5 |
|---|---|---|
| Success | 100% | 0% |
| Median time | 1.8 s | 55 s |
| Cost per run | $0 | $0.052 |
9 runs each, 7 November 2026. How we measure
What your agent gets
Two recipes are published for Lieferando. Each one shows up as a tool in your agent through wield-mcp.
| Recipe | What it does | Login | Status (7 Nov 2026) |
|---|---|---|---|
lieferando.de/search-restaurants | Lists restaurants that deliver to, or offer pickup near, an address | No | OK. 9 of 9 benchmark runs passed |
lieferando.de/order-status | Shows your recent orders: restaurant, order time, status, delivery estimate and total | Yes | Untested. Needs a logged-in recording |
The search takes an address and optional filters: service (delivery or collection), cuisine, name, open now, sort and a result limit of 5, 10, 25 or 50. Each restaurant comes with its rating, delivery time range, distance, delivery fee, minimum order, free-delivery threshold, cuisines and deals. Paid sponsored placements are removed by default, so the order follows Lieferando's organic ranking. The how-to page has the parameters and example prompts.
How a request flows
- You ask your agent: "Find restaurants that deliver to my address in Berlin and list the cuisines."
- The agent calls
find_recipe, thenrunwithlieferando.de/search-restaurantsand your address. - wield-mcp passes the run through a local bridge to the Wield extension.
- The extension sends the request from a lieferando.de tab, so it carries your browser session.
- The results go back to your agent. Nothing passes through wield.run.
What leaves your computer
The search sends its requests to one host, rest.api.eu-central-1.production.jet-external.com. Wield does not follow redirects to other hosts. Passwords and cookies never leave your computer. A search takes three requests. Wield logs each request on your computer only, with the method, host, path, status and timing. It never logs bodies or query values.
Lieferando terms, in plain words
Lieferando's terms prohibit scraping and automated systematic retrieval. The recipe is built for one search at a time that you start yourself, from your own browser. It is not for bulk work. The address lookup is billed to Lieferando by Google, so do not use the recipe for bulk geocoding. Ordering stays on lieferando.de. Check the site's terms before you automate anything else.
Limits and what is not working yet
- Order status is untested. The recipe needs a logged-in recording before anyone should rely on it.
- No ordering. Neither recipe places an order. Ordering stays on lieferando.de.
- Bot checks stop the run. When a site shows a check, such as a CAPTCHA or an "are you human" page, Wield stops and asks you to solve it in your own tab. Wield does not solve bot checks. In our benchmark, bot protection blocked the cloud browser agent in all nine runs.
FAQ
Is there an official Lieferando MCP server?
Not that we know of as of 7 November 2026. Wield provides this search as a recipe that runs in your browser.
Which AI apps work with it?
Any MCP client. npx -y wield-mcp init sets up Claude Desktop, Claude Code, Cursor and Codex. The AI apps page lists each one.
Does it need a login?
The restaurant search does not. The order status recipe does, and it is untested as of 7 November 2026.
How fast is it compared with a browser agent?
In our benchmark, as of 7 Nov 2026 (preliminary, still running), the recipe passed 9 of 9 runs with a median of 1.8 seconds. Browser Use driving Claude Sonnet 5.5 passed 0 of 9 runs, with a median of 54.8 seconds and about $0.05 per run. Bot protection blocked the browser agent in all nine runs. The benchmark page has the method and raw logs.