Skip to main content
Glama
welt-mcp

welt-mcp

README.md
# welt-mcp 📰

mcp-name: io.github.mohammad-abbasi-sh/welt-mcp

Unofficial **read-only** MCP server for **[WELT](https://www.welt.de)** — one of Germany's biggest daily newspapers — via its official RSS feed at [www.welt.de/feeds/latest.rss](https://www.welt.de/feeds/latest.rss) and article pages, whose full text is read from the JSON-LD `articleBody`. Top headlines, news by topic (Deutschland, Ausland, Fußball, Kultur, …), keyword search, the latest video items, and full article text. No key, no login.

> Built by Mohammad Abbasi as part of the "Germany MCP factory" (Sepehr-style pipeline). Not affiliated with WELT or Axel Springer.

## For AI agents

Add to your MCP client config (stdio):

```json
{
  "mcpServers": {
    "welt": {
      "command": "uvx",
      "args": ["welt-mcp==0.1.0"]
    }
  }
}
```

Or run locally: `python -m welt_mcp.server` (needs `mcp<2`, `httpx`).

## Tools

| Tool | Description |
|------|-------------|
| `welt_headlines` | Top headlines (official RSS feed, 30 items), newest first — `limit` 1..50 |
| `welt_news_by_topic` | Newest news of one topic — `topic`: "Deutschland", "Ausland", "Fußball", "Kultur", "Film", "Debatte", "Eishockey", … (case-insensitive substring match), `limit` 1..50 |
| `welt_search_news` | Search news by keyword (title + teaser) in the feed — `query` (e.g. "Bundestag"), `limit` 1..50 |
| `welt_latest_videos` | Latest video items from the feed (subtype "video"), newest first — `limit` 1..50 |
| `welt_get_article` | Full text of one WELT article page (JSON-LD `articleBody`) — `url`: a www.welt.de article URL (containing /article or /plus). WELT+ articles return `paywalled: true` with the teaser only |
| `welt_health` | Reachability probe (ok / blocked / error) — one polite request |

Example output (`welt_headlines`, `limit` 2):

```json
{"feed_title": "Nachrichten - WELT", "total_results": 30, "count": 2,
 "items": [
  {"title": "Baugenehmigungen in Deutschland brechen im September deutlich ein – doch Jahresbilanz bleibt positiv",
   "link": "https://www.welt.de/politik/deutschland/article6afd6d12fc384ca056385c40/baugenehmigungen-in-deutschland-brechen-im-september-deutlich-ein-doch-jahresbilanz-bleibt-positiv.html",
   "description": "Der Aufwärtstrend im deutschen Wohnungsbau ist gestoppt: Die Zahl der genehmigten Wohnungen sank im September um knapp 15 Prozent. Höhere Zinsen und Baukosten belasten die Branche.",
   "author": "Till Henniges", "published": "2026-11-18T10:10:03+00:00", "category": "Deutschland",
   "topic": "Jahresvergleich", "premium": false, "subtype": "news",
   "image": "https://images.welt.de/6afd6d12fc384ca056385c40/771566a806d7370faa5b330a76bd385b-master/ci23x11-w1200/dpa-com-261030-935-1707967-jpg"},
  ...]}
```

Notes:
- The server reads the **official RSS feed** (30 items, verified live 2026-11-18). Search and topic filters run client-side over the fetched feed. Each item carries the feed's own `<welt:premium>` flag and `<welt:subType>` (news / video / opinion).
- **Paywall honesty:** WELT+ (premium) articles expose only their teaser in the JSON-LD — `articleBody` equals the description (~200 chars, verified on 3 premium items). `welt_get_article` returns `paywalled: true` with the teaser for those, never fabricated text. Free articles return the full body (1,500–3,900 chars in the sampled items). Short agency flashes can also be teaser-only (`teaser_only: true`, `paywalled: false`) — the teaser is then the whole freely available text.
- Section feeds do not exist: `/feeds/<section>.rss` is 404 for every section tried (politik, wirtschaft, sport, kultur, …); the single `latest.rss` feed is the only feed, and topics are its `<category>` values.
- Data: WELT public RSS feed and article pages — no key, no login, read-only.

## Development

```bash
python3 -m venv .venv && .venv/bin/pip install "mcp<2" httpx pytest
.venv/bin/pip install -e .
.venv/bin/python -m pytest tests/ -q
```

- `tests/test_domain.py` — offline, parses the real captured RSS feed + one free and one WELT+ article page
- `tests/test_live.py` — cautious live requests against www.welt.de

Tests: 41/41 green (34 offline + 7 live, 2026-11-18).

## License

[PolyForm Noncommercial 1.0.0](https://polyformproject.org/licenses/noncommercial/1.0.0/) — free for personal and non-commercial use; commercial use needs a separate license (see LICENSE).

---

Built by [Mohammad Abbasi](https://github.com/mohammad-abbasi-sh) — [Almanibama](https://almanibama.ir) · [mabbasi.ir](https://mabbasi.ir)

TDQS

A3.9/5.0

Scored across 6 tools

Disambiguation4/5

welt_get_article and welt_health are clearly distinct, but four tools (headlines, news_by_topic, search_news, latest_videos) all filter the same 30-item RSS feed and could be confused at the edges — e.g. headlines already contains video and category items. The descriptions do clarify the differences (topic filter vs. keyword search vs. subtype filter), so selection is workable but not perfectly crisp.

Naming Consistency4/5

All names share the welt_ prefix and snake_case, which makes them scannable and cohesive. There is a minor structural deviation between verb-first names (welt_get_article, welt_search_news) and noun-phrase names (welt_headlines, welt_latest_videos, welt_health), but the pattern remains predictable enough.

Tool Count5/5

Six tools is well-scoped for a read-only news source: one detail fetch, four feed views (headline, topic, search, video) and one health probe. Each tool earns its place with no filler or redundant surface.

Completeness4/5

The read path is fully covered: discover items, filter by topic or keyword, grab full article text, and check feed availability. Gaps are minor but real — search only spans the 30-item feed with no pagination or date-range query, and there is no tool to enumerate valid topic/category names, forcing agents to guess from the description examples.

Maintenance

ActivityMaintained
ResponsivenessOutdated data