MCP.so
登录

Claude Custom Prompts MCP Server

@minipuft

基本信息

分类

AI 与智能体

许可证

AGPL-3.0

传输方式

stdio

作者

minipuft

关于 Claude Custom Prompts MCP Server

MCP prompt template server: hot-reload, thinking frameworks, quality gates

Claude Custom Prompts MCP Server 配置

使用下面的配置,将此服务器添加到你的 MCP 客户端。

{
  "mcpServers": {
    "claude-prompts-mcp": {
      "command": "npx",
      "args": [
        "-y",
        "claude-prompts@latest",
        "--init=~/my-prompts"
      ]
    }
  }
}

Claude Custom Prompts MCP Server 工具

工具列表暂未提供

当前收录暂时没有可展示的工具列表,不代表服务没有工具。请查阅项目文档了解支持的能力。

点击上方按钮,在临时沙箱中运行此服务器并实时获取工具列表。

概览

What is 🚀 Claude Custom Prompts MCP Server?

It is an MCP workflow server that lets you craft reusable prompts with validation and reasoning guidance, orchestrate multi‑step agentic workflows using a composable operator syntax, and export them as native skills. Designed for AI clients like Claude Code, Claude Desktop, VS Code, Cursor, OpenCode, and Gemini CLI, it provides 90+ bundled prompt templates, gates for output validation, and hot‑reloadable YAML resources.

How to use 🚀 Claude Custom Prompts MCP Server?

Install via Claude Code marketplace (/plugin install claude-prompts@minipuft) for hooks and bundled resources, or run with npx (npx -y claude-prompts@latest --client=claude-code) for Claude Desktop and other editors. Invoke prompts with the >> syntax (e.g., >>tech_evaluation_chain library:'zod' context:'API validation'). Optionally create your own workspace with --init=~/my-prompts and set MCP_WORKSPACE or MCP_RESOURCES_PATH to load custom prompts alongside the defaults.

Key features of 🚀 Claude Custom Prompts MCP Server

  • Versioned YAML prompt templates with hot‑reload
  • Validation gates that block or advise on output quality
  • Composable workflow operators (chain, gate, agent handoff)
  • Export any workflow as a native skill for the client
  • 90+ bundled prompts across 11 categories

Use cases of 🚀 Claude Custom Prompts MCP Server

  • Multi‑step research chains that thread context through gates
  • Code review workflows with automated output validation
  • Self‑correcting agentic sequences that retry failed gates
  • Custom skill authoring and export to any supported client

常见问题

1

What are Validation Gates?

Gates are criteria the AI checks its own output against. They can be blocking (must pass before continuing) or advisory (flag issues but allow proceed). Failed gates can trigger automatic retries or pause for your decision.

2

Which AI clients are supported?

Claude Code, Claude Desktop, VS Code (Copilot), Cursor, OpenCode, and Gemini CLI. Each has specific setup instructions and may have different levels of hook support (chain tracking, gate enforcement).

3

How do I use my own prompts instead of the bundled ones?

Run `npx -y claude-prompts@latest --init=~/my-prompts` to create a starter workspace, then set the `MCP_WORKSPACE` environment variable. Your prompts overlay the bundled ones; same‑ID resources take priority.

4

Does it require Node.js and npm?

It depends on the transport. The NPX method requires Node.js, but a self‑contained `.mcpb` bundle (~5MB) is available for Claude Desktop that requires no npm.

5

What transport options are available?

The default transport is `stdio`. For HTTP clients you can use `--transport=streamable-http`. Presets per client are set via the `--client` flag (e.g., `claude-code`, `gemini`).

评论

AI 与智能体 分类下的更多 MCP 服务器