Crossfiled / API

The API

Everything Crossfiled knows, readable by software. No key, no sign-up, no rate card. There is an OpenAPI spec and an MCP server if you would rather point a tool at it than read this.

Cite the newsroom, not us. Crossfiled hosts no journalism. Every item carries the URL of the outlet that filed it, and that is who wrote it. Cite Crossfiled only for the thing we actually produce: the observation that several newsrooms covered one event, and the order in which they filed.

Getting started

curl https://crossfiled.com/api/v1/stories?min_outlets=3&limit=5

No authentication. 120 requests a minute per IP. Every response is CORS-open and cached for two minutes. If you need more than that, write to kevin@champlinenterprises.com rather than working around it.

Endpoints

GET/api/v1/stories

Current stories, newest and widest-covered first.

`topic`	One of us-politics, world, courts-justice, money, tech-ai, health, climate
`outlet`	Only stories this outlet covered, e.g. Reuters
`min_outlets`	At least this many outlets. Use 3+ to drop single-outlet items
`since`	ISO date or datetime
`has_video`	true to require video coverage
`include_opinion`	true to include opinion outlets (off by default)
`limit`	1 to 100, default 25

GET/api/v1/stories/{id}

One event with every outlet version attached, in filing order.

GET/api/v1/sources

Every outlet we read, with its format class and country.

GET/api/v1/topics

Topics and how many current stories each holds.

GET/story/{id}/{slug}

The human page. Returns JSON with Accept: application/json.

GET/feed.json

The raw ingest output, if you would rather do your own grouping.

What a story looks like

{
  "id": "c_89b02bef06",
  "title": "Networks suspend pool coverage after White House bars CNN",
  "url": "https://crossfiled.com/story/c_89b02bef06/...",
  "outlet_count": 13,
  "outlets": ["BBC News", "CNN", "NPR", "..."],
  "video_count": 7,
  "article_count": 21,
  "first_filed": "2026-09-19T14:02:11Z",
  "last_filed":  "2026-09-21T18:44:02Z"
}

outlet_count is the one worth paying attention to. It is a measure of how widely an event is being reported, and on a typical day only about 25 of 500 stories are carried by more than one newsroom.

MCP server

An assistant can query Crossfiled directly rather than scraping it. Stateless JSON-RPC 2.0 at POST /mcp, no auth.

{
  "mcpServers": {
    "crossfiled": { "type": "http", "url": "https://crossfiled.com/mcp" }
  }
}

`search_stories`	Filter by topic, outlet, breadth of coverage, video, date
`get_story`	Every outlet's version of one event, in filing order
`list_sources`	Every outlet, with its format class
`coverage_spread`	Which outlets covered an event and which did not

coverage_spread is the one no single outlet can answer. Read a silence as an observation and not a verdict: an outlet may be working a different angle, filing later, or simply not serving that beat.

Also