This server lets MCP-capable AI assistants browse, read, and search structured daily AI news from AI Daily Insights without HTML scraping.
Get the latest issue with all its structured news items in one call (get_latest).
List recent issues with date, title, summary, tags, and URLs (list_latest).
Fetch a specific issue by date, with news items parsed into index, title, signal, and body (get_article).
Retrieve every issue in an inclusive date range, fetched in parallel (get_range).
List issues by tag (company/topic), matched case-insensitively (list_by_tag).
Search across all news items by keyword, tag, or both, with snippets (search).
All tools validate inputs, return structured data, and need no API key.
Click on "Deploy Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type @ followed by the MCP server name and your instructions, e.g., "@ai-daily-insights-mcp List the latest 5 AI news items"
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
A Model Context Protocol server for AI Daily Insights — a daily AI briefing built for both humans and agents.
It lets any MCP-capable assistant (Claude Desktop, Claude Code, Cursor, …) list, read and search the daily AI news, fully structured — no HTML scraping.
Each entered tool handler sends a best-effort background event to the configured AI_DAILY_BASE_URL (the public site by default). It contains only the tool name and package version: no arguments, search queries, results, session IDs or user identifiers. The hosting infrastructure still sees standard request metadata. Events include cached calls and errors, but exclude tool discovery and rejected input schemas. They are self-reported usage, not verified AI users.
Set AI_DAILY_ANALYTICS=0 in the MCP server environment to disable these events. Data fetches still use the package/version User-Agent for HTTP attribution. Reporting never waits on the tool response path, has a 1.5-second timeout, at most 16 outstanding events and no retries; offline/shutdown/opt-out may undercount. Existing data caching is unchanged.
Related MCP server: News API MCP Server
Tool | What it does |
| The newest issue with all its news items in one call. |
| List recent issues: date, title, summary, tags, URLs. |
| Fetch one issue ( |
| Every issue in a date range (e.g. the past week), fetched in parallel. |
| Issues carrying a given tag (company/topic), matched case-insensitively. |
| Query financing reports, with evidence and amount qualifiers. |
| Company timeline, oldest first; exact company name or slug by default. |
| Full-text + tag search across all news items, in one server-side call. |
Data comes from the public JSON endpoints (/index.json, /{date}.json, /search) — the server is stateless and needs no API key. Responses are cached in-memory briefly (default 5 min, override with AI_DAILY_CACHE_TTL_MS) to avoid hammering the site on repeated calls.
Financing tools (v0.5.0+) read /financing/deals.json. The dataset is CC BY 4.0: credit AI Daily Insights and link the deal URL. Preserve evidence, notes and corrections. Report dates are not announcement dates; repeated coverage is not a new round; debt/credit facilities are not equity or cash received.
claude mcp add ai-daily-insights -- npx -y ai-daily-insights-mcp{
"mcpServers": {
"ai-daily-insights": {
"command": "npx",
"args": ["-y", "ai-daily-insights-mcp"]
}
}
}Then ask: "用 ai-daily-insights 列出最近 5 天的 AI 资讯，挑 SpaceX 相关的讲讲。"
Env var | Default | Purpose |
|
| Override the site origin (e.g. a preview deploy). |
|
| In-memory cache TTL for fetched JSON, in milliseconds. |
npm install
npm test
npm run inspect # opens the MCP InspectorRequires Node.js 20 or newer. Every tool declares explicit MCP safety annotations, validates its inputs with Zod, and converts upstream failures into structured MCP error results.
MIT
Fetch a single daily issue by date, parsed into structured news items: each item has index, title, signal (the one-line judgment) and body. Date format: YYYY-MM-DD.
Inputs: date (required).
Fetch the most recent issue, parsed into structured news items {index, title, signal, body}. Use this when the user asks for today's / the latest AI news and does not give a date.
Inputs: no named parameters.
Fetch every issue whose date falls within [from, to] (inclusive), each parsed into structured news items. Use for "the past week", "between X and Y", etc. Dates are YYYY-MM-DD. Issues are fetched in parallel.
Inputs: to (required), from (required), limit.
List issues that carry a given tag (e.g. a company or topic like "OpenAI", "融资", "芯片"). Returns issue-level entries; use search() to find specific news items by keyword.
Inputs: tag (required), limit.
List the most recent AI Daily Insights issues (date, title, one-line summary, tags, URLs). Start here to discover what is available.
Inputs: limit.
Full-text + tag search across all news items. Provide a keyword query, a tag, or both. Returns matching items with date, title, signal and a short snippet. Backed by the site search endpoint (one call); falls back to the flat index if unavailable.
Inputs: tag, limit, query.
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
Scored across 6 tools
Each tool has a distinct purpose: fetching by date, latest, range, listing by tag, listing latest issues, and full-text search. There is minor overlap between list_latest and get_latest, and between list_by_tag and search, but descriptions clarify the differences (issue-level vs. item-level).
All names use lowercase snake_case and follow a verb-based structure. However, the pattern is mixed: get_article, get_latest, get_range are consistent; list_by_tag and list_latest use 'list', and search is a bare verb. This is minor inconsistency overall.
With 6 tools, the count is well-scoped for a news retrieval service. Each tool covers a specific access pattern (by date, latest, range, tag, list, search), and none feel redundant or out of place.
The tool surface covers the main ways to retrieve articles: by date, latest, range, tag listing, and full-text search. A notable gap is the lack of a tool to list all available tags, which would assist in using list_by_tag and search effectively.
Give your AI assistant access to the world’s news. Search articles in any language, track coverage over time, and research companies and events.
Real-time curated crypto news for AI agents with sentiment, recaps, and search.
Real-time financial news for AI agents: search by ticker and source, with sentiment and entities.
AI papers, dev-tool adoption, AI news and startup funding rounds, updated daily. No key needed.
