❯ What is AI Daily Insights?

AI Daily Insights is a daily AI-industry briefing built for both human readers and AI agents. The same reporting is published as pages a person can skim and as endpoints a machine can discover, cite and search.

The goal is simple: a human opening the page gets judgement, context and a signal; an agent calling the site gets stable URLs, a structured index and explicit citation boundaries.

❯ Three ways in — pick one

Route	Who it's for	How to start
MCP (recommended)	Claude / Cursor and other AI assistants	One command to register; let the assistant read for you (below)
CLI	Terminal people	`npx ai-daily-insights latest`
JSON / RSS	Developers / custom agents	`GET /index.json`, `/feed.xml`

❯ What can you do once it's connected?

With MCP connected you don't need to remember any commands — just talk to your assistant in plain language. It picks the right tool and the right day on its own. Two common patterns:

📰 Daily skim

Show me today's AI Daily
What were the big AI stories in the last three days?
Summarise this week's AI Daily into 5 bullets

🔍 Company / topic tracking

Search recent news about Anthropic
What did OpenAI ship this week? Give me the details
Keep an eye on Nvidia and HBM memory

Behind the scenes those plain-language asks map to tools automatically:

What you want	How to ask	Tool called
Recent issues	“Any recent AI news?”	`list_latest`
Today / latest issue	“Show me today's AI Daily”	`get_latest`
One specific day	“Open the June 24 issue”	`get_article`
A date range / this week	“Big stories in the last three days”	`get_range`
Everything on a topic	“List every issue tagged OpenAI”	`list_by_tag`
Full-text / tag search	“Search for news about Anthropic”	`search`

❯ Why “built for humans and agents”?

Traditional media pages assume a human reader; APIs assume a program. This site tries to make the layer between them thinner: every issue exists at once as HTML, RSS, a JSON index and llms.txt, so browsers, feed readers, search engines and agents can all enter from whichever door suits them.

❯ How should an AI agent call this site?

Three steps: read /llms.txt to understand the structure, read /index.json for the list of issues with summaries, then open individual article URLs for the full text. Every endpoint is a public GET — no auth, no API key.

# 1. Discovery: site guide + endpoint list
curl -s https://www.aidailyinsights.cn/llms.txt

# 2. Index: date / title / summary / URL for every issue
curl -s https://www.aidailyinsights.cn/index.json

# 3. Full text: open one day's article (HTML)
curl -s https://www.aidailyinsights.cn/2026-06-24/

If you use Claude, ChatGPT or any assistant with tools, hand it this and it will walk the three steps itself:

Read https://www.aidailyinsights.cn/index.json,
list the titles and summaries of the last 5 days of AI news;
when I pick one, open its article URL and walk me through the details.

English readers can start from the English endpoints instead:

curl -s https://www.aidailyinsights.cn/en/feed.xml            # English RSS
curl -s https://www.aidailyinsights.cn/en/search-index.json  # English flat index

❯ One-command MCP setup (Claude / Cursor)

There is an official MCP server, ai-daily-insights-mcp, published on npm. Once installed, your assistant can list, read and search the briefing without being fed URLs by hand.

Claude Code — one command:

claude mcp add ai-daily-insights -- npx -y ai-daily-insights-mcp

Claude Desktop / Cursor — add to the config file:

{
  "mcpServers": {
    "ai-daily-insights": {
      "command": "npx",
      "args": ["-y", "ai-daily-insights-mcp"]
    }
  }
}

Six tools are exposed:

Tool	What it does
`get_latest`	The newest issue with every story, in one call.
`list_latest`	The last N issues (date / title / summary / tags / URL).
`get_article`	One issue by date, parsed into `{index, title, signal, body}`.
`get_range`	Every issue in a date range (e.g. the past week), fetched in parallel.
`list_by_tag`	Issues carrying a tag (company / topic); tags are case-insensitive.
`search`	Site-wide full-text + tag search, answered by `/search` in one call.

 Responses are cached in memory for a few minutes (5 by default, tunable via AI_DAILY_CACHE_TTL_MS), so follow-up questions don't hammer the site. Source and docs: npm / GitHub.
 ❯ Reading it in the terminal (CLI)
 If you prefer the command line, the ai-daily-insights CLI reads the briefing without a browser:
 npx ai-daily-insights latest          # last 5 issues
npx ai-daily-insights show 2026-06-24 # one issue, every story + signal
npx ai-daily-insights search Anthropic # search across issues
 It uses the same public JSON endpoints as MCP; no login. Source: npm / GitHub.
 ❯ Agent calling protocol
  Discovery: /llms.txt — site guide and endpoint list for agents.
 Index: /index.json — date, title, summary and article URL for every issue.
 Structured issue: /{date}.json — one issue split into {index, title, signal, body} items.
 Search: /search?q=…&tag=… — full-text + tag search in one call; /search-index.json is the flat index of every item.
 English: /en/feed.xml and /en/search-index.json mirror the above for the English edition.
 Subscribe: /feed.xml — RSS, suitable for polling.
 Citation: keep the article URL, publication date and title; do not treat the one-line summary as the full factual source.
 
  signal For an agent, this site is not a look-first content library but a set of news nodes meant to be discovered, cited and recombined.
 
 ❯ About the English edition
 Every issue is written in Chinese first and translated to English automatically, keeping the same slug and the same story order. The Chinese page is the canonical original — each English article links back to it, and both are cross-declared via hreflang. If a rendering of a fact looks off, check the Chinese original before citing.