SuperAPI AI Directory

ai-daily-insights-mcp

List, fetch and full-text search structured daily AI news from AI Daily Insights, a Chinese AI-industry briefing.

Open website

About this tool

Purpose: To provide a Model Context Protocol server for the AI Daily Insights (aidailyinsights.cn) daily AI briefing, enabling any MCP-capable assistant to list, read, and search daily AI news, with fully structured data and no HTML scraping required.

Main capabilities: Provides three types of tools—list, fetch, and full-text search—with data coming from public JSON endpoints (/index.json, /{date}.json, /search). The server is stateless, requires no API key, and responses are briefly cached in memory (5 minutes by default, overridable with AI_DAILY_CACHE_TTL_MS). Each tool declares explicit MCP safety annotations, uses Zod to validate input, and converts upstream failures into structured MCP error results.

Usage: Requires Node.js 20 or later. After configuring this MCP server in the configuration file of Claude Desktop or Cursor, you can ask questions such as "Use ai-daily-insights to list the AI news from the last 5 days, and pick out SpaceX-related items to talk about."

Privacy and aggregate usage (0.4.4+): Each incoming tool handler makes a best effort to send a background event to the configured AI_DAILY_BASE_URL (by default the public site), containing only the tool name and package version, and not containing parameters, search queries, results, session IDs, or user identifiers; hosted infrastructure may still see standard request metadata. Events include cache calls and errors, but not tool discovery or rejected input schemas, and represent self-reported usage rather than verified AI users. This can be disabled by setting AI_DAILY_ANALYTICS=0 in the MCP server environment. Data fetching still uses the package/version User-Agent for HTTP attribution. Reporting does not block the tool response path, has a 1.5-second timeout, allows at most 16 in-flight events, and does not retry; being offline, shut down, or exited may result in undercounting. Existing data caching is unchanged.

Limitations: The materials do not provide specific tool names, parameter details, installation commands, or complete configuration examples; before use, confirm the Node.js version and compatibility with the MCP client.