News Search
1 creditGET /api/premium/news/searchThe news search endpoint runs a full-text search over the article corpus with relevance scoring (term hits in title weighted 3x over snippet 1x, plus a recency boost). Filters: date range, provider substring match, category. Stop words are stripped. Each result includes title, url, source, snippet, published_at, relevance score, and the matched query terms.
When to use this endpoint
When an agent needs targeted search ("what did Anthropic publish in March?", "any news mentioning agent payments?"). For a curated daily summary instead of search, see whats-new.
Parameters
| Name | In | Type | Description |
|---|---|---|---|
| q | query | string | Free-text query. Omit to browse latest filtered articles. |
| from | query | string | Start date YYYY-MM-DD UTC |
| to | query | string | End date YYYY-MM-DD UTC |
| provider | query | string | Substring match against source name and domain |
| category | query | string | Substring match against article categories |
| limit | query | integer | Max results (1-100, default 25) |
* required
Example response
{
"ok": true,
"query": "opus pricing",
"matched": 12, "returned": 12,
"results": [
{
"title": "Anthropic ships Claude Opus 4.7", "url": "...",
"source": "Anthropic Blog", "published_at": "2026-04-27T12:00:00Z",
"relevance": 0.85, "matched_terms": ["opus"]
}
]
}Code samples
Python SDK
from tensorfeed import TensorFeed
tf = TensorFeed(token="tf_live_...")
r = tf.news_search(q="agent payments", from_date="2026-04-01")TypeScript SDK
import { TensorFeed } from 'tensorfeed';
const tf = new TensorFeed({ token: 'tf_live_...' });
const r = await tf.newsSearch({ q: 'agent payments' });MCP tool
Available via the TensorFeed MCP server as news_search. Add npx -y @tensorfeed/mcp-server to your Claude Desktop or Claude Code MCP config.
FAQ
How does the relevance score work?
A token in the title contributes 3 points; a token in the snippet contributes 1 point. The total is normalized by query length so longer queries do not automatically out-score shorter ones (caps at 0.7). Then a recency boost in [0, 0.3] is added based on how recent the article is in the matched window. Final score is in [0, 1].