dw-mcp
# dw-mcp ๐ฐ
mcp-name: io.github.mohammad-abbasi-sh/dw-mcp
Unofficial **read-only** MCP server for **[dw.com](https://www.dw.com)** โ Deutsche Welle news in 31 languages from DW's official sitemaps: latest headlines, keyword search over recent headlines, latest articles, full article text and topic pages. No key, no login. No key, no login.
> Built by Mohammad Abbasi as part of the "Germany MCP factory". Not affiliated with dw.com.
## For AI agents
Add to your MCP client config (stdio):
```json
{
"mcpServers": {
"dw": {
"command": "uvx",
"args": ["dw-mcp==0.1.0"]
}
}
}
```
Or run locally: `python -m dw_mcp.server` (needs `mcp<2`, `httpx`).
## Tools
| Tool | Description |
|------|-------------|
| `dw_headlines` | Latest DW headlines โ `language` (e.g. "en", "de", "fa-ir"), `limit` 1..50 |
| `dw_search_news` | Keyword search over the latest DW headlines โ `query`, `language`, `limit` 1..50 |
| `dw_latest_articles` | Latest DW articles by publication time โ `language`, `limit` 1..50 |
| `dw_get_article` | Full text of one DW article โ `url` (a dw.com article link) |
| `dw_topics` | DW topic pages matching a word โ `query`, `language`, `limit` 1..50 |
| `dw_health` | Reachability probe (ok / blocked / error) โ one polite request |
## Notes
- All data comes from **DW's official sitemaps** (free, no key, no login).
- DW has **no working RSS feeds any more** โ every `/rss/` path returns 404 (verified 2026-11-10) โ and its site search is gone, so headlines come from the news sitemaps and `dw_search_news` filters the latest ~50 headlines (not the full archive).
- Persian is two editions: `fa-ir` (Iran) and `fa-af` (Afghanistan/Dari); plain `fa` does not exist.
- Article bodies are extracted from DW's server-rendered pages (JSON-LD headline/date + cleaned `<p>` text); live blogs and video pages may return little or no body text.
Tests: 22/22 green (15 offline + 7 live, 2026-11-10).
## 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
```
## License
MIT
---
Built by [Mohammad Abbasi](https://github.com/mohammad-abbasi-sh) โ [Almanibama](https://almanibama.ir) ยท [mabbasi.ir](https://mabbasi.ir)
TDQS
Scored across 6 tools
dw_get_article, dw_topics, and dw_health are clearly distinct, but dw_headlines, dw_latest_articles, and dw_search_news all return recent items in a language and differ mainly in source (news sitemap vs article feed vs keyword filter). The descriptions do draw those lines explicitly (search is limited to the ~50 sitemap items), so confusion is limited but not zero.
All names share a dw_ prefix and snake_case with a predictable noun/verb structure (dw_get_article, dw_search_news, dw_latest_articles). dw_headlines, dw_topics, and dw_health omit the verb, which is a minor deviation rather than an inconsistency.
Six tools is well-scoped for a news-reading server: listing, searching, fetching full text, browsing topics, and a connectivity probe. Each tool has a clear, non-redundant job with no filler.
The core read path (list headlines/articles โ fetch full article) is covered, but there are notable dead ends: dw_topics returns topic pages with no tool to list articles within a topic, and search is limited to the ~50-item sitemap with no date-range or archive option. A listing-by-topic or date-filtered search would round out the surface.