EclipseWatcher: Using Agents and MCP to Find the latest Berserk Chapter Drops

What is EclipseWatcher?

EclipseWatcher is an MCP server that gives LLM agents the ability to search the web for news about the manga Berserk. It scrapes the chapter list from readberserk.com using a Cloudflare-aware scraper and exposes tools for function-calling agents, designed for readers who want automated updates.

How to use EclipseWatcher?

Clone the repository, start the VLLM server with start_server.sh, then run the code blocks inside qwen_client.ipynb. The server exposes two MCP tools: get_latest_berserk_chapter() and is_new_berserk_chapter(last_seen), which agents can call to answer questions about the newest chapter.

Key features of EclipseWatcher

• Cloudflare-aware scraper using cloudscraper and BeautifulSoup
• MCP tools auto-described for function-calling LLM agents
• Supports get_latest_berserk_chapter() and is_new_berserk_chapter() queries
• One-line integration with Qwen-client via VLLM
• FastMCP-based server for easy agent integration

Use cases of EclipseWatcher

• Ask an LLM agent “What’s the newest chapter of Berserk?” and get the answer from live data
• Automate notifications when a new Berserk chapter drops by polling the is_new_berserk_chapter tool
• Integrate Berserk chapter tracking into a larger agent workflow (e.g., daily summaries)

FAQ from EclipseWatcher

What runtime dependencies does EclipseWatcher require?

VLLM, FastMCP, requests, beautifulsoup4, websockets, and asyncio, plus Python 3.12+ (inferred from badge).

Where does the data come from?

EclipseWatcher scrapes the Berserk chapter list directly from readberserk.com.

Does EclipseWatcher handle Cloudflare protection?

Yes, the scraper is built with cloudscraper to bypass Cloudflare challenges, but site changes may break the scraper.

What transport or protocol does EclipseWatcher use?

It exposes tools via the Model Context Protocol (MCP) using FastMCP, making it compatible with any MCP client.

Does EclipseWatcher require authentication?

No authentication is mentioned; it scrapes a public website and runs locally without external API keys.