For AI assistants

Read the paper through your AI

Connect The Morning Science to Claude, ChatGPT, Cursor or any other assistant that speaks MCP, and ask it about the news in your own words. It gets the stories as data – briefs, sources, discussion – instead of guessing from a web search, and it cites the original reporting.

Server address

https://themorning.science/api/mcp

Free, read-only, no account, no key.

What you can ask

“What’s in the paper today? Just the security stuff.”
Today’s front page or any past edition, filtered by section, each story with its brief and the original source.
get_edition
“What did I miss since yesterday?”
Only what is new or got new coverage. The assistant keeps a cursor, so next time it picks up exactly where it left off.
what_changed_since
“Here’s my package.json – does any of this week’s news affect me?”
Security advisories naming your exact packages (with vulnerable and patched versions), releases with breaking changes and outages of what you use. Works great in Claude Code or Cursor, where the assistant can read the file itself.
check_my_stack
“Has anything happened with Postgres lately?”
Searches 30 days of news – headlines, briefs, companies, projects, versions and CVE ids.
search_stories
“Tell me more about that OpenSSH bug – what do developers say?”
One story in full: every outlet that covered it, the timeline, and what science readers said on Hacker News, Lobsters & co.
get_story

You don’t pick the tool – the assistant does. It also chains them: “Check my dependencies, then tell me which of the hits is the most urgent and what developers say about it” is one question.

How to connect it

Claude (web and desktop)

Settings → Connectors → Add custom connector. Name it “The Morning Science”, paste the server address, done. Turn it on in a chat from the tools menu.

Claude Code

claude mcp add --transport http themorningscience https://themorning.science/api/mcp

ChatGPT

Settings → Apps & Connectors → Advanced settings → turn on Developer mode, then create a connector with the server address and no authentication. Custom connectors depend on your ChatGPT plan.

Cursor, VS Code, Windsurf and others

Add it to the editor’s MCP config (for Cursor: .cursor/mcp.json):

{
  "mcpServers": {
    "themorningscience": {
      "url": "https://themorning.science/api/mcp"
    }
  }
}

What the assistant gets

Every story comes with the headline, the source and a link to the original article, our story page, and – once the story has been briefed – a few lines of facts, why it matters, what to do, the versions and CVEs involved, how many outlets covered it and how much it was discussed. Only stories that passed the news check are in it, the same as in the paper.

The briefs are written by AI from the reporting they cite, as on the story pages. Headlines and articles belong to their publishers – an assistant that answers from The Morning Science should link them, and the data makes that easy.

No MCP? Plain files work too

Every story page as Markdown: add .md to its address. Handy for pasting into a chat or for agents that browse.
/llms.txt – a map of the site and today’s news for assistants and answer engines.
RSS for the front page and every section.

What we see

The server only answers questions; it doesn’t know who is asking and has no login. What your assistant sends – a search term, a list of packages – is used to answer that one request and not stored. Your conversation with the assistant never reaches us, only the tool calls it decides to make. See the privacy policy for the server logs every request leaves.

For developers

Streamable HTTP transport, stateless, JSON responses (no SSE stream), protocol versions 2024-11-05 to 2025-11-25. All tools are read-only and return structuredContent plus the same data as text. CORS is open. Try it:

curl -s https://themorning.science/api/mcp \
  -H 'content-type: application/json' \
  -d '{"jsonrpc":"2.0","id":1,"method":"tools/call",
       "params":{"name":"check_my_stack","arguments":{"dependencies":["next","axios"]}}}'