Zaviye · Developers

Multilingual news clusters, API-accessible.

The same pipeline that publishes the Zaviye newsroom is exposed as a commercial API. Farsi, Kurdish, Arabic, Turkish, and English-language press — ingested, clustered by embedding similarity, cross-language-matched, framing-annotated, and served as JSON. REST and MCP interfaces. Self-serve billing. No sales calls.

Read the docs Get a key Status Tiers

What Zaviye gives you that other news APIs don't

Farsi / Kurdish / Arabic / Turkish native ingestion

227 feeds across 8 languages, indexed as first-class corpora — not an afterthought. Most commercial news APIs offer English-only or bolt-on translation of English sources.

Cross-language clustering

Clusters are matched across corpora by embedding similarity. A story that ran in Kurdish and Turkish but not English is one cluster with three language slots, not three separate records.

Framing + bias annotations

Per-cluster framing analysis in each corpus, source-spectrum breakdown, state-media flags. The analysis layer runs on every cluster by default on paid tiers.

Coverage-gap signals

Endpoints that return clusters with high inter-corpus disparity — the stories one side filled pages on and another barely touched. A sellable primitive for media-monitoring products.

MCP + REST from the same data

Build a Claude integration with the MCP server today; move to REST for production traffic; same cluster store, same account, same rate bucket.

Same data the Zaviye newsroom publishes

No separate pipeline, no separate quality tier. What the API returns is what readers see on zaviye.news.

Shape of a call

$ curl https://zaviye.news/api/v1/feed \
    -H "Authorization: Bearer zv_live_••••" \
    --data-urlencode "lang=fa,ku,ar,he,tr,en" \
    --data-urlencode "limit=50" \
    -G

{
  "clusters": [
    {
      "id": "clu_8Q3…",
      "headline": "A regional rail corridor …",
      "article_count": 18,
      "languages": { "fa": 9, "ku": 5, "tr": 3, "ar": 1, "en": 1 },
      "blind_spot": true,
      "blind_spot_corpus": "en",
      "ai_summary_short": "…",
      "framing": { "fa": "…", "ku": "…", "tr": "…", "en": "…" },
      "sources": [ … ],
      "updated_at": "2026-04-20T09:04:12Z"
    }
  ],
  "usage": { "calls_this_period": 1243, "quota": 50000 }
}

MCP Quickstart · Claude Desktop

Zaviye exposes a full MCP server via HTTP. Point any MCP-compatible client at the endpoint — the OAuth 2.0 + PKCE flow runs automatically.

// claude_desktop_config.json
{
  "mcpServers": {
    "zaviye": {
      "url": "https://zaviye.news/api/mcp",
      "transport": "http"
    }
  }
}

// Available tools: news_feed · cluster_detail · translate_cluster · profile

Starter tier: MCP read scope included free. Insider+: full scope with framing annotations and higher rate limits.

Tiers

All self-serve · except Enterprise

Starter

Free

Rate: 1,000 calls / month
Window: 24-hour data window
Languages: English corpus
Keys: 1 API key

For prototyping, Claude integrations, and seeing what the shape of the data is before you build on it.

Includes

Access to /api/v1/feed and /api/v1/cluster/:id
MCP endpoint via OAuth
Changelog and status page

Typical buyers

· Claude / LLM personal integrations
· Journalism student projects
· Evaluating whether Zaviye data fits your app

Self-serve · Paddle billing

Subscriber

Recommended

$9 / month

$90 yearly · save two months

Rate: 50,000 calls / month
Window: 7-day data window
Languages: Farsi · Kurdish · Arabic · Turkish · English
Keys: 1 API key

The same Subscriber plan on /pricing. One subscription covers personal reading and personal API use — 50k calls/month included.

Includes

All Starter endpoints
Full five-language cluster access
Translation pairs (title + summary per language)
Search by source language corpus
Consumer: Letter archive, Globe view, quarterly zine

Typical buyers

· Indie developers building news-adjacent apps
· Small research projects
· Single-person newsletters that want clusters
· Claude / LLM power users wanting more than the free tier

Start Subscriber

Self-serve · Paddle billing

Insider · Studio

$25 / month

$250 yearly · for small teams up to 5 seats

Rate: 500,000 calls / month
Window: 30-day data window
Languages: All five + framing + bias annotations
Keys: Up to 5 API keys

The Insider plan ($25/mo on /pricing), scaled for small teams. Multi-seat API access with framing analysis and coverage-gap signals.

Includes

All Subscriber endpoints
Framing analysis per cluster per corpus
Source-bias breakdowns (spectrum + state-media flags)
Coverage-gap signals (which corpora missed which stories)
Webhook support (basic)
Up to 5 consumer seats
48-hour support SLO

Typical buyers

· Media monitoring firms
· Small analytics products
· Language-services agencies
· Think-tank research programmes

Start Insider · Studio

Self-serve · Paddle billing

Insider · Press

$60 / month

$600 yearly · for teams up to 100 seats

Rate: Unlimited reasonable use
Window: Full historical archive
Languages: All five + full annotation layer
Keys: Up to 25 API keys

For newsrooms, research institutes, and media-intelligence firms whose workflows need continuous access at an institutional price.

Includes

Everything in Insider · Studio
Full historical archive (back to project start)
Priority webhook delivery
25 API keys for team access
Up to 100 consumer seats
48-hour support SLO
Uptime status page visibility

Typical buyers

· Newsrooms with bureau operations
· Academic departments
· Media-intelligence firms
· Large translation services
· AI companies training on multilingual corpora

Start Insider · Press

Self-serve · Paddle billing

Enterprise / Site Licence

From $15,000 / year

Annual invoice / PO

Rate: Dedicated rate bucket
Window: Full archive + historical dumps
Languages: Everything + negotiable custom
Keys: Unlimited

For multinational media groups, LLM labs licensing the corpus, and institutions whose procurement process requires a signed order form.

Includes

Everything in Insider
Unlimited consumer seats (Site Licence)
Historical archive dumps (CSV / Parquet)
Single annual invoice (30-day net)
Dedicated rate bucket (noisy neighbour immunity)
Custom SLA options
Acceptable-use consultation

Typical buyers

· Multinational media groups
· LLM labs licensing news corpora
· Government-adjacent research institutions
· Large philanthropic foundations funding a research programme

Request a quote

Short form · invoice within two business days

What no tier includes

Custom endpoint development
Private clustering on your own RSS feeds
White-label or co-branded editions
Streaming / WebSocket feeds (REST + MCP only)
Write endpoints (no posting to Zaviye via API)
GraphQL
Phone or video support at any tier
SLAs below Enterprise
Dedicated support engineers
Reselling per-source RSS relays

Every absence is a product decision. If you need any of the above, Zaviye is politely not the right API.

Acceptable use — the short version

What you can do

Build commercial products on Zaviye data, including paid end-user products
Display cluster data, framing analysis, and bias breakdowns in your UI
Cache responses within the data window of your tier
Train models on response data, credited to Zaviye (Press / Enterprise only)
Combine Zaviye data with other sources

What you can't

Resell raw article URLs as if they were a feed you own
Ship a product that looks like a Zaviye clone (brand imitation)
Use Zaviye data to generate disinformation, targeted harassment, or engagement-bait content
Share your API key with parties outside your organisation
Circumvent rate limits via multi-account registration

Full policy lives at /developers/aup (coming). Keys may be revoked for policy violations with 30-day notice except for acute misuse, where revocation is immediate.

Enterprise — request a quote

Tell us what you're building. We'll send pricing and an invoice within two business days.

No sales call. No discovery meeting. We read the form, we price it, you decide. If your procurement department needs a signed order form, include their email and we'll correspond directly.