---
name: wondel-mcps
description: Live data through 26 free, authless remote MCP servers at mcp.wondel.ai (Streamable HTTP, no key, no signup). Use when an answer needs current prices, listings, fares or facts instead of training knowledge, above all for Poland and prices in PLN - flights (Google Flights, Wizz Air, Polish charters, Skyscanner, Kayak, FlightAware status), places to stay (Google Hotels, Booking, Airbnb, noclegi.pl, AlohaCamp, Slowhop), Google Maps places and directions, Google web search, YouTube videos and transcripts, shopping and price comparison (Allegro, Ceneo, OLX, Amazon, eBay, Google Shopping), NBP exchange rates and gold, Polish property listings and transaction prices (Otodom, Gratka, Morizon, nieruchomosci-online, deweloperuch). Covers connecting a client, picking the server with the /mcp directory, and calling politely (latency, caching, one query per question, guided errors).
---

# Wondel.ai MCP servers

26 remote MCP servers on one host, each wrapping one public site or official API. No API key, no signup, no OAuth: an MCP client connects and calls tools at once. Data is fetched live from the source for each query, so every result is as current as the site behind it.

## When to use

Reach for a server when the user needs something you cannot know: today's fare, a listing that exists now, a shop's price, an exchange rate, what a place or video says. One user question maps to one or two tool calls on one server.

Leave them alone for background knowledge and for bulk work: these servers answer a person's own lookup, not a crawl. The legal notice at the end of this file is the boundary.

## Connect

Every server is a Streamable HTTP endpoint at `https://mcp.wondel.ai/<id>/mcp`, where `<id>` is the slug from the list in this file. Send no `Authorization` header and no key.

Claude Code, one server per command:

```bash
claude mcp add --transport http otodom https://mcp.wondel.ai/otodom/mcp
```

Any other MCP client, in its servers config (the transport key name varies by client; the URL is all a server needs):

```json
{
  "mcpServers": {
    "otodom": { "type": "http", "url": "https://mcp.wondel.ai/otodom/mcp" }
  }
}
```

`https://mcp.wondel.ai/<id>/sse` answers HTTP 410: the 2024 HTTP+SSE transport is gone, so a client that only speaks SSE cannot connect.

## Pick a server

The list below is usually enough. To see tool names before connecting, connect to the directory server at `https://mcp.wondel.ai/mcp` and call `list_servers`, filtered by `category` or `query` (a substring over title, summary and tool names); each entry carries its `endpoint`, `status`, `latencyNote` and `tools`.

The slugs are the `<id>` in the endpoint URL. "experimental" marks a server whose source blocks it more often; prefer a stable neighbour in the same category when one covers the question.

### Shopping

- `allegro`: Live offers on Allegro.pl: search, browse a category, fetch one offer. Latency: ~30–120 s, live stealth-browser scrape via unblocker chain; results with offers cached 10 min.
- `ceneo`: Price comparison on Ceneo.pl: search products, list every shop's offer, fetch product details. Latency: ~3–20 s, HTML scrape via unblocker chain (forced in prod).
- `google-shopping`: Google Shopping product search: listings with prices, merchants and ratings. Latency: ~2–8 s via SerpAPI.
- `ebay`: Search any eBay marketplace and fetch full item details for one listing. Latency: ~2–8 s via SerpAPI.
- `olx`: OLX.pl classifieds in every category — property to rent or buy, cars, electronics, jobs, free items — plus offer details. Latency: <1 s, direct fetch of OLX's public JSON API (Firecrawl/Bright Data fallback only on a block).
- `amazon`: Amazon product search on 22 country stores, multi-store price comparison and one product's offers, every price in PLN (NBP rates). Latency: ~1.5–4 s per search via SerpAPI; amazon_compare runs its stores in parallel (1 credit each); amazon_product ~10–25 s.

### Travel

- `google-flights`: Live Google Flights fares: search routes, discover cheap deals from an origin. Latency: ~2–8 s via SerpAPI.
- `google-hotels`: Live Google Hotels search: properties with rates and ratings, plus photos and reviews. Latency: ~2–8 s via SerpAPI.
- `wizzair`: Wizzair fares, routes, timetables and flight details across its network. Latency: ~2–10 s, direct fetch to be.wizzair.com; wizzair_search_flights is experimental (Kasada-protected endpoint).
- `czartery`: Polish charter flight search across Rainbow, Itaka and TUI operators, prices in PLN. Latency: ~2–10 s, direct fetch to each operator's JSON/RSC endpoint.
- `skyscanner` (experimental): Live Skyscanner fares in PLN: best-ranked one-way or round-trip itineraries, cheapest/fastest price, booking links. Latency: ~20–40 s, live render of the skyscanner.pl results page via Firecrawl; 5 min cache.
- `booking` (experimental): Experimental Booking.com hotel search: nightly price, rating, stars and booking link. Latency: ~3–20 s, HTML scrape; AWS WAF often forces the unblocker fallback.
- `flightaware`: Live flight status and scheduled-route search via the official FlightAware AeroAPI. Latency: ~1–3 s, official AeroAPI; needs FLIGHTAWARE_API_KEY, bills usage per call.
- `kayak`: Builds a Kayak.com flight-search deep link for a route and dates — URL only, no prices. Latency: instant — pure URL builder, no HTTP call at all.
- `noclegi`: Places to stay in Poland on noclegi.pl: hotels, guest houses, apartments, cottages, with total prices for your dates. Latency: ~1–2 s, direct HTML fetch + the site's place autocomplete (unblocker fallback only on a blocked page).
- `alohacamp`: Outdoor stays on AlohaCamp.com: cabins, glamping, tiny houses and camping pitches in Poland, with dated prices in PLN. Latency: ~1–2 s, direct fetch of the site's Next.js JSON routes and API (unblocker fallback only when blocked).
- `airbnb` (experimental): Airbnb stays worldwide: search by place, dates and guests with total prices, ratings and links, plus one listing's details. Latency: ~1–2 s, direct fetch of airbnb.pl pages (Firecrawl fallback only on a challenge page, ~8 s); listing price = 2 requests.
- `slowhop`: Slow stays on Slowhop.com: cabins, agritourism, glamping and guesthouses in nature, mostly in Poland, priced for your dates, party and pets (PLN). Latency: <1 s, the site's own JSON API (direct; ZenRows fallback only on a blocked response).

### Places

- `google-maps`: Google Maps places: local search, place details, reviews, and turn-by-turn directions. Latency: ~2–8 s via SerpAPI.

### Search

- `youtube`: YouTube search: find videos, fetch details and comments, pull full transcripts. Latency: ~2–8 s via SerpAPI.
- `google-search`: Fast Google web search: organic results and related questions, no ads. Latency: ~1–4 s via SerpAPI (google_light engine).

### Money & rates

- `nbp`: Official Polish central bank (NBP) exchange rates, currency conversion and gold prices. Latency: <1 s, free public api.nbp.pl (no key).

### Real estate

- `nieruchomosci-online`: Polish property listings on nieruchomosci-online.pl: flats, houses, plots, premises for sale or rent. Latency: ~1–3 s, direct HTML fetch (unblocker fallback only on a blocked page).
- `gratka`: Polish property listings on Gratka.pl and Morizon.pl (one shared database): flats, houses, plots, premises for sale or rent, with details. Latency: <1 s, the site's own GraphQL API (direct; ZenRows fallback only on a blocked response).
- `otodom`: Polish property listings on Otodom.pl: flats, houses, rooms, plots, commercial for sale or rent, plus full offer details. Latency: ~0.3–1 s, direct fetch of the site's Next.js JSON routes (page HTML via unblocker only when blocked, ~6 s).
- `deweloperuch`: Polish flat prices from deweloperuch.pl: RCN transaction prices (valuation, city statistics, rankings, buildings) and developers' asking prices. Latency: ~1–5 s, direct fetch of deweloperuch.pl's JSON API, at most one request in flight and one per second per isolate — per isolate, not global (a call makes 1–2 requests, each waiting its turn for up to 20 s); results cached 1–24 h; Warsaw's offer dump is 1.3 MB; stops on a block (401/403/429/challenge).

## Call politely

- **Expect the latency the list states for each server,** and wait that long before treating a call as failed.
- **Results are cached.** Each server keeps results for minutes to a day. An identical call inside that window returns the cached answer, so repeating a call does not fetch fresher data.
- **One query per question.** Every call reaches a real site, and many servers spend shared paid provider credits per call. Ask the specific question the user asked; a loop over dates, cities or pages, or a fan-out of hundreds of calls, is bulk querying, which the terms forbid when it rebuilds a source's data; keep to the user's own question.
- **Read a failed result, then fix the input.** A tool failure travels inside an ordinary HTTP 200 MCP result with `isError: true` and text that starts `Error:`. When the call itself is at fault, the text names the parameter or value to change and ends with `See https://mcps.wondel.ai/servers/<id>/`; change what it names or open that page before calling again, because the same call repeats the same error. A provider outage reads as a plain `Error:` with nothing to fix: wait and retry once, later.
- **Keep the source in the answer.** Results carry the source's name and original links; pass both on, and tell the user to confirm price and availability at the source before paying or booking.

## Legal notice
- Unofficial, experimental servers run by Wondel.ai, free of charge; not affiliated with or endorsed by any source site.
- Data is fetched on demand from third-party sites and official APIs. Copyright, database rights and trademarks stay with the source operators; nothing here grants a licence.
- Intended for personal use (looking something up for yourself through an assistant), plus testing, research and demonstration of MCP integrations. Not for bulk querying, redistribution, competing databases or training datasets.
- Attribute results to their source and keep the original links. Data may be stale or wrong; verify at the source before deciding. Provided as-is, without warranty.
- Rights holders and site operators: write to support@wondel.ai and the server is withdrawn promptly (it then answers HTTP 451). The same address takes reports of inaccurate data and personal-data removal requests.
- Full terms, Polish binding with an English translation: https://mcp.wondel.ai/legal.txt

## More

- Every tool with its parameters: https://mcp.wondel.ai/llms-full.txt
- Routing index: https://mcp.wondel.ai/catalog.json; AI Catalog: https://mcp.wondel.ai/.well-known/ai-catalog.json
- Per-server docs for humans: https://mcps.wondel.ai