AI导哈

MCP工具集

这是一个通过 Homebrew 分发的 MCP 工具集，提供 Web 内容读取、Zed 聊天记录搜索和任务管理等功能，适合 JavaScript 渲染网站与 AI 上下文高效处理场景。

  • by mredig
  • Ruby
  • 本地部署
  • 更新于
在 GitHub 查看 mredig/homebrew-pizza-mcp-tools

使用说明

Pizza MCP Tools

Homebrew tap for MCP (Model Context Protocol) tools.

Why These Tools?

Modern web applications rely heavily on JavaScript rendering, which can be challenging for basic HTTP fetch approaches. This tap provides specialized web tools designed to handle these scenarios, offering improvements over some built-in MCP client capabilities:

  • JavaScript rendering - Executes JavaScript and waits for DOM stability, ensuring complete content extraction from SPAs (React, Vue, etc.)
  • Smart pagination - Fetch large pages in manageable chunks (default: 2,500 chars) to preserve AI context
  • Intelligent caching - Dramatically faster repeated requests
  • Integrated search - Search within pages and across the web with multiple engines
  • Symbiotic workflow - Tools share cache and work together: discover URLs → read content → search within

Real-world impact: Sites like Google Search that might yield 166 characters with basic fetch return 7,345 characters with JS rendering. Reddit returns 16KB of actual content instead of 3KB of skeleton HTML.

Installation

Add this tap:

brew tap mredig/pizza-mcp-tools

Available Formulae

mcp-webreader

Swift-based MCP server for web content reading, designed for modern JavaScript-heavy web applications and large webpages that destroy ai contexts.

Install:

brew install mcp-webreader

Features:

  • JavaScript rendering - WKWebView with DOM stability detection for dynamic sites
  • Three specialized tools:
    • fetch-page - Fetch and paginate web content
    • search-page - Search within webpages with character position results
    • search-web - Multi-engine web search (Google, DuckDuckGo, Bing, Brave)
  • Smart pagination - Default 2,500 char chunks, customizable offset/limit
  • Shared caching - All tools benefit from cached content
  • HTTP customization - Custom headers, user agents, any HTTP method
  • Link extraction - With optional same-site filtering

Performance:

  • Fresh fetch (no JS): ~0.3s
  • Fresh fetch (with JS): ~2.6s
  • Cached fetch: ~0.04s (7-26x faster)

Repository: https://github.com/mredig/MCP-WebReader

mcp-zedhist

Swift-based MCP server and CLI for searching and accessing Zed editor agent chat history. One formula installs two binaries: mcp-zedhist (the MCP server) and zedhist (the terminal CLI).

Install:

brew install mcp-zedhist

MCP features:

  • Five specialized tools:
    • zed-list-threads - List threads with optional date filtering and limits
    • zed-get-message - Get specific message with character-level pagination
    • zed-get-metadata - Get thread/message metadata (counts, roles, timestamps, model info)
    • zed-search-threads - Search thread titles/summaries
    • zed-search-thread-content - Search within messages with limited context
  • Token-efficient - Returns snippets with context, not full conversations
  • Smart caching - Caches decompressed thread content (zstd compression)
  • Message indexing - Access messages by index for stable retrieval
  • Date filtering - Filter threads by update date range

CLI features (zedhist):

  • zedhist list-threads - List threads with date/limit filters
  • zedhist get-message <threadID> - Print a specific message
  • zedhist extract-json-blobs - Dump decompressed thread JSON to a folder
  • zedhist troubleshoot - Sanity-check the threads database

Performance:

  • Handles searches across 299 threads with 9K+ messages
  • Cache eliminates repeated decompression (~1.7s for all threads)
  • String matching instead of regex for fast searches

Repository: https://github.com/mredig/MCP-ZedChat

fingerstring

Swift-based CLI tool for task list management with support for subtasks, notes, and list organization.

Install:

brew install fingerstring

Features:

  • Task management - Create, edit, delete, and complete tasks
  • Subtasks - Organize tasks hierarchically with unlimited nesting
  • Notes - Add detailed notes to tasks and lists
  • List organization - Create and manage multiple task lists
  • CLI commands:
    • list-create, list-view, list-delete, list-all - List management
    • task-add, task-view, task-edit, task-delete, task-complete - Task management
  • Persistent storage - Tasks stored in SQLite database

Repository: https://github.com/mredig/FingerString

mcp-fingerstring

Swift-based MCP server exposing FingerString task management as an MCP interface. LLMs can use the mcp and users can track via the cli tool externally from the llm.

Install:

brew install mcp-fingerstring

Features:

  • Task tools:
    • task-view - View task details and subtasks
    • task-edit - Edit task label or note
    • task-add - Add tasks to lists or as subtasks
    • task-delete - Delete tasks
    • task-complete - Mark tasks complete/incomplete
  • List tools:
    • list-view - View list with recursive subtask display
    • list-create - Create new task lists
    • list-delete - Delete lists
    • list-all - Show all lists

Repository: https://github.com/mredig/MCP-FingerString

Usage

After installation, the tools are available as executables:

fingerstring --help
mcp-webreader --help
mcp-zedhist --help
zedhist --help
mcp-fingerstring --help

Refer to each tool's repository for detailed usage instructions and MCP integration.

相关 MCP