Readgzh Site

What

ReadGZH is a server-side proxy service that converts WeChat Official Account articles into clean, AI-readable formats. It targets developers, researchers, and AI users who need to ingest Chinese web content that is typically blocked or obfuscated for external bots. By stripping proprietary HTML tags and styles, it reduces token consumption and improves parsing accuracy for LLMs.

Features

• Server-Side Proxy & Cleaning: Fetches WeChat articles via a backend proxy, removing inline styles, private tags (e.g., <mp-common-profile>), and nested containers to output pure semantic HTML or Markdown.
• MCP Server Integration: Includes a native Model Context Protocol (MCP) endpoint (POST https://api.readgzh.site/mcp-server), allowing AI agents to call the service directly via standardized protocol without custom API wrappers.
• Shared Cache System: Articles are stored in a shared cache; subsequent reads by any user or agent do not consume credits, reducing costs for high-frequency or multi-agent workflows.
• CDN Image Proxying: Routes article images through a CDN to prevent broken links caused by WeChat’s short-lived image URLs, ensuring visual context remains accessible to AI vision models.
• Long-Form Chunking & Summarization: Supports splitting long articles into ~40KB chunks via ?part=N for stable ingestion, and offers a Pro-only ?mode=summary endpoint for structured abstracts.

Helpful Tips

• Use Markdown Output: When using the API, append &format=text to the request URL. This returns Markdown, which is generally more token-efficient and structurally clearer for LLMs than raw HTML.
• Leverage the MCP Endpoint: If your AI agent framework supports MCP, connect directly to the ReadGZH MCP server. This eliminates the need to manually construct HTTP requests and handles authentication/protocol nuances automatically.
• Monitor Cache Status: Check if an article is already in the shared cache before making a request. Cached reads are free, which is critical for cost management when processing large batches of historical articles.
• Handle Edge Cases: Be prepared for explicit error responses on deleted, video-only, or authorization-restricted articles. Implement fallback logic in your pipeline to skip or flag these items rather than failing the entire batch.

OpenClaw Skills

ReadGZH’s native MCP support makes it a strong candidate for integration into the OpenClaw ecosystem as a "Content Ingestion" skill. A likely use case involves building an automated research agent that monitors specific WeChat Official Accounts, uses ReadGZH to fetch and clean new posts, and then summarizes or archives them into a knowledge base. This combination could streamline competitive intelligence or market research workflows for teams tracking Chinese digital media, reducing the manual effort required to bypass WeChat’s access restrictions. This is a plausible workflow based on the provided MCP and API capabilities, not a confirmed pre-built OpenClaw plugin.