返回 MCP 目录
public公开dns本地运行

mcp记事摘要

这个基于 TypeScript 的服务器实现了一个简单的笔记系统,允许用户创建和管理文本笔记并生成摘要,展示了核心的 MCP 概念。

article

README

systemprompt-agent-server

npm version smithery badge License: Apache-2.0 Twitter Follow Discord

Website | Documentation | Blog | Get API Key

A specialized Model Context Protocol (MCP) server that enables you to create, manage, and extend AI agents through a powerful prompt and tool management system. This server integrates with systemprompt.io to provide seamless creation, management, and versioning of system prompts through MCP.

An API KEY is required to use this server. This is currently free, although this may change in the future. You can get one here.

This server uses Sampling and Notification functionality from the @modelcontextprotocol/sdk. This will only work with advanced clients that support these features. A free opensource client multimodal-mcp-client can be used to provide a complete voice-powered AI workflow solution.

Required Client

This server is designed to work with the multimodal-mcp-client - a voice-powered MCP client that provides the frontend interface. Please make sure to set up both components for the full functionality.

Why Use This Server?

  • Agent Management: Create and manage AI agents with customized system prompts and tool configurations
  • Extensible Tool System: Add, modify, and combine tools to enhance your agents' capabilities through MCP
  • Prompt Management: Centralized management of system prompts with versioning and metadata support
  • Type-Safe Integration: Full TypeScript support with proper error handling
  • MCP Compatibility: Works seamlessly with multimodal-mcp-client and other MCP-compatible clients
  • Open Source: Free to use and modify under the MIT license

Features

Core Functionality

  • MCP Protocol Integration: Full implementation of Model Context Protocol for seamless AI agent interactions
  • Voice-Powered Interface: Compatible with voice commands through multimodal-mcp-client
  • Real-Time Processing: Supports streaming responses and real-time interactions
  • Type-Safe Implementation: Full TypeScript support with proper error handling

Agent Management

  • Create and configure AI agents with specific capabilities
  • Manage agent states and contexts
  • Define agent behaviors through system prompts
  • Monitor and debug agent interactions
  • Version control for agent configurations
  • Resource management for agent assets

Advanced Tools System

Built-in tools include:

  • Prompt Management
    • create_prompt - Create new system prompts with metadata
    • edit_prompt - Update existing system prompts with versioning
    • get_prompt - Retrieve specific prompt configurations
  • Resource Management
    • create_resource - Create new agent resources and configurations
    • edit_resource - Modify existing agent resources
    • list_resources - Browse available agent resources
    • read_resource - Access specific agent resource content
  • System Tools
    • systemprompt_heartbeat - Monitor system status and health
    • systemprompt_fetch_resources - Retrieve all available resources
  • Agent Management
    • create_agent - Create new systemprompt agents
    • edit_agent - Modify existing systemprompt agents
    • list_agents - View available systemprompt agents

Sampling & Notifications

  • Advanced sampling capabilities for AI responses
  • Real-time notification system for agent events
  • Configurable sampling parameters
  • Event-driven architecture for notifications

Integration Features

  • API Key management and authentication
  • User status and billing information tracking
  • Subscription management
  • Usage monitoring and analytics

Development Tools

  • Built-in debugging capabilities
  • Test utilities and fixtures
  • Type-safe mocking utilities
  • Comprehensive testing framework

🎥 Demo & Showcase

Watch our video demonstration to see Systemprompt MCP Client in action:

▶️ Watch Demo Video

The demo showcases:

  • Voice-controlled AI interactions
  • Multimodal input processing
  • Tool execution and workflow automation
  • Real-time voice synthesis

Development

Install dependencies:

npm install

Build the server:

npm run build

For development with auto-rebuild:

npm run watch

Installation

Installing via Smithery

To install SystemPrompt Agent for Claude Desktop automatically via Smithery:

npx -y @smithery/cli install systemprompt-agent-server --client claude

Manual Installation

To manually configure with Claude Desktop, add the server config:

On MacOS: ~/Library/Application Support/Claude/claude_desktop_config.json On Windows: %APPDATA%/Claude/claude_desktop_config.json

{
  "mcpServers": {
    "systemprompt-agent-server": {
      "command": "/path/to/systemprompt-agent-server/build/index.js"
    }
  }
}

Debugging

Since MCP servers communicate over stdio, debugging can be challenging. We recommend using the MCP Inspector, which is available as a package script:

npm run inspector

The Inspector will provide a URL to access debugging tools in your browser.

Testing

This project uses Jest for testing with TypeScript and ESM (ECMAScript Modules) support.

Test Configuration

The test setup includes:

  • Full TypeScript support with ESM modules
  • Global fetch mocking
  • Automatic test reset between runs
  • Custom matchers for validation
  • Type-safe mocking utilities

Module Resolution

The project uses a dual module resolution strategy:

  • Source code uses ESM (ECMAScript Modules) with .js extensions
  • Tests use CommonJS for compatibility with Jest

This is configured through two TypeScript configurations:

  • tsconfig.json: Main configuration for source code (ESM)
  • tsconfig.test.json: Test-specific configuration (CommonJS)
// Source code imports (ESM)
import { Something } from "../path/to/module.js";

// Test file imports (CommonJS)
import { Something } from "../path/to/module";

Running Tests

# Run tests
npm test

# Watch mode
npm run test:watch

# Coverage report
npm run test:coverage

Test Structure

Tests are located in __tests__ directories next to the files they test. The naming convention is *.test.ts.

Related Links

help

运行方式说明

cloud

托管运行

托管运行通常表示这个 MCP Server 由服务方环境承载,用户一般按页面提供的连接方式或授权流程接入,不需要在本地长期启动一个 MCP 进程

  1. 打开服务方连接页
  2. 完成授权或复制端点
  3. 在 MCP 客户端中连接
terminal

本地运行 / 其它方式

本地运行通常需要用户在自己的电脑或服务器上安装依赖,把 server_config 复制到 MCP 客户端,并按 env_schema 补齐环境变量、密钥或其它配置

  1. 复制 server_config
  2. 安装所需依赖
  3. 补齐环境变量后重启客户端