Multilingual news API
A multilingual news API should not hand you the same event once per language. CLSTR folds coverage in many languages into one event, writes its title and summary in English, and links up to 100 original sources per event in the language each was published in, with the outlet count, over REST and MCP.
The problem: one event in twelve languages is twelve results
A presidential runoff is covered at home in Portuguese, across the region in Spanish, in French by francophone Africa, and in Arabic, Persian, Hindi and Vietnamese. Per-language feeds give you twelve streams to merge. A translate-then-search pipeline gives you one stream, a translation bill per article and a matching problem: two translated headlines about the same event rarely share their words.
What you want is the event once, with the evidence of how far it travelled.
How clusters cross languages
CLSTR reads about 107,000 articles a day from about 20,000 sources a day, in 58 languages and from publishers in 176 countries, over the 30 days to October 21, 2026. It groups coverage by meaning, using multilingual embeddings, not by shared words, so a French, an Arabic and a Portuguese write-up of the same result usually land in one cluster, with a source count, a significance score from 1 to 10 generated by an AI model, and its source articles. Related clusters link into a situation with a timeline; how that works is on News clustering API.
What you get back
- Title and summary in English. Written by a model from the coverage, so one query language reaches every region.
- The original sources, untranslated. Each article comes back with
source_host,url,published_atand its title as published. - The real outlet count.
articles_totalis the number of outlets behind the event. The list holds up to 100 representative articles, one per source, andarticles_truncatedsays when it was cut.
Worked example: one election, many languages
Search in English, then open the event with the most outlets.
curl -s "https://api.clstr.news/v1/search?q=Brazil%20election&days=7" \ -H "Authorization: Bearer clstr_YOUR_KEY" curl -s "https://api.clstr.news/v1/clusters/b552e38f-979d-459e-aefe-c7f3eb39d20d?articles=100" \ -H "Authorization: Bearer clstr_YOUR_KEY"
Over MCP, as tools/call params:
{ "name": "search_situations",
"arguments": { "query": "Brazil election" } }
{ "name": "get_cluster",
"arguments": { "cluster_id": "b552e38f-979d-459e-aefe-c7f3eb39d20d", "articles": 100 } }
Captured 2026-10-28 via get_cluster (trimmed to seven of the 100 articles returned):
{
"title": "Lula da Silva re-elected President of Brazil in historic second-round victory",
"summary": "Luiz Inácio Lula da Silva has won Brazil's presidential election, defeating Flávio Bolsonaro in a historic second-round comeback to secure a fourth term.",
"sources": 2264,
"articles_total": 2264,
"articles_truncated": true,
"articles": [
{ "source_host": "acidadenoticia.com.br",
"title": "Lula é reeleito presidente do Brasil em virada histórica" },
{ "source_host": "diariosalta.com",
"title": "Lula da Silva ganó un cuarto mandato en Brasil contra Bolsonaro" },
{ "source_host": "auxnouvelles.tg",
"title": "Brésil : Lula da Silva réélu pour un 4e mandat" },
{ "source_host": "alnaharps.com",
"title": "لولا دا سيلفا يفوز بولاية رئاسية رابعة في البرازيل بعد منافسة حادة مع فلافيو بولسونارو" },
{ "source_host": "baolaocai.vn",
"title": "Tổng thống Brazil đắc cử nhiệm kỳ thứ 4" },
{ "source_host": "dobrudjabg.com",
"title": "Лула да Силва - новият стар президент на Бразилия" },
{ "source_host": "banjaluka.net",
"title": "Brazilski predsjednik Lula osvojio četvrti mandat" }
]
}
One English summary stands over 2,264 outlets. In the 100 articles returned, titles were in Portuguese, Spanish, French, Arabic, Persian, Hindi, Bengali, Vietnamese, Bulgarian, Greek, Italian, Albanian, Indonesian, Bosnian and English. Those languages are read off the original titles, returned as published.
If this event is no longer available when you read this, search for any election in the last week and open the result with the highest sources.
Coverage, measured
The language and country counts above were measured over the 30 days to October 21, 2026. The observatory breaks down what the pipeline reads by language, over its own window. Look up the languages you need there to see how much of the corpus each one is. To narrow by place rather than language, use the country filter on News API by country.
How events are formed
- Multi-outlet, across languages. An event surfaces once more than one outlet is covering it, in any language CLSTR reads, so you get stories the press is reporting, not single-source noise.
- English out, originals kept. Titles and summaries are written in English; each source keeps its title as published, with host, link and time.
- It shows what the press reports. A summary says what the coverage says, drawn from outlets writing in many languages.
- History windows. Listing and search reach back 7 days on Free (100 requests and 10 searches a day), 30 days on Builder (5,000 and 250).
How it compares
Categories, not named products; individual products vary.
| CLSTR | Translate-then-search pipelines | Per-language feeds | |
|---|---|---|---|
| Same event across languages | Usually one cluster, with linked sources and the outlet count | Matching is your job | One stream per language |
| Output language | English titles and summaries; original titles kept | Whatever you translate into | The original language |
| Translation on your side | None | Per article | None, but no merge |
| Outlet count | Per event | You count it | Per feed |
Start
- Connect an assistant or framework: MCP server for news.
- Choosing a source for an agent: News API for AI agents.
- What the pipeline reads, by language: the observatory.
- Compare tiers: pricing.
- Endpoints, fields and caps: API and MCP reference.