Newsroom API

Checked news, by API.

Ask for a story. Get it back as JSON or a signed webhook, with its sources and check result. Held stories never reach your app.

Copies a setup prompt for your AI coding tool

Base URL: https://api.tlsnewsroom.com

Quickstart

Three calls to your first story.

  1. 1Get an API keyCreate one in the Console under Integrations. Send it as a Bearer token.
  2. 2Ask for a storyPOST to /generations with a story type. Add an Idempotency-Key so retries never double-bill.
  3. 3Read what passedGET /articles for ready stories, or add a webhook and get each one as it lands.
Ask for a story, then read it
NR=https://api.tlsnewsroom.com/v1/newsrooms/nr_markets

# Ask your newsroom for a story
curl -X POST $NR/generations \
  -H "Authorization: Bearer $NEWSROOM_API_KEY" \
  -H "Idempotency-Key: fed-decision-0917" \
  -H "Content-Type: application/json" \
  -d '{ "output_format": "ticker_card" }'

# Then read the stories that passed
curl $NR/articles \
  -H "Authorization: Bearer $NEWSROOM_API_KEY"
Build it with AI

Or let your coding tool do it.

One prompt sets up Newsroom in your project. Copy it into any AI coding tool, or open it straight in Claude, ChatGPT, Cursor, or GitHub Copilot.

Copies a setup prompt for your AI coding tool

Copy the setup prompt, or click a tool to open it there. It builds the client, the feed, and the webhook for you.

  • Asks for your API key and keeps it on the server
  • Builds a small client for the Newsroom API
  • Adds a news feed that shows the AI label
  • Sets up a webhook and checks its signature
The response

Every story comes with its proof.

Only stories that passed are in /articles. Here’s what each one carries.

title, summaryThe headline and a one-line summary
body_markdownThe full story, in Markdown
source_urlsLinks to every source it used
disclosureThe AI label to show your readers
gate_reportWhat the check found, fact by fact
evidence_idPull the full evidence from /evidence
{
  "data": [{
    "status": "ready",
    "format": "ticker_card",
    "title": "Fed holds at 4.25% — fourth straight pause",
    "summary": "Futures ticked up 0.3% after the statement.",
    "disclosure": "Written by an AI writer",
    "source_urls": ["https://www.federalreserve.gov/…"],
    "gate_report": { … }
  }]
}
Endpoints

The calls you’ll use most.

Every route needs your API key. Responses come wrapped in { data, meta }.

Stories

POST/v1/newsrooms/{id}/generationsAsk for a story
GET/v1/newsrooms/{id}/generations/{generation_id}Check on a request
GET/v1/newsrooms/{id}/articlesStories that passed
GET/v1/newsrooms/{id}/articles/{article_id}/evidenceThe sources behind a story
POST/v1/newsrooms/{id}/previewPreview, never billed

Your newsroom

POST/v1/newsroomsCreate a newsroom
POST/v1/newsrooms/{id}/sourcesAdd a source
POST/v1/newsrooms/{id}/sources/{source_id}:testTest a source
POST/v1/newsrooms/{id}/authorsAdd a writer
POST/v1/newsrooms/{id}/webhooksAdd a webhook
Story types

Set output_format.

ticker_cardTicker, 35–80 words, $0.20
feed_itemFeed item, 120–250 words, $0.30
briefBrief, 400–800 words, $0.50
analysisAnalysis, 900–1,500 words, $10
fullFull story, 1,500–2,500 words, $30
Webhooks

Signed, and retried.

Each delivery has a Newsroom-Signature header. Check it with your webhook secret, like the Webhook tab above.

article.readyA story passed and is in your feed
generation.blockedA story was held. Not billed
article.retractedYou pulled a story back
generation.failedSomething broke. Not billed
For AI assistants

Your newsroom, inside your assistant.

The same newsroom and the same check, from the AI tools your team already uses. Until then, the API does it all.

MCP serverComing soon

Ask for checked stories from Claude, Cursor, or any MCP client. Your sources, your rules, the same check.

  • Ask your AI assistant for a story on your beat
  • Read your feed and each story’s sources in chat
  • Same check, same price, held stories still free

Build with checked news.

Get a key, ask for a story, and see the sources behind it.