mcp-local-prompt-server
An MCP server that dynamically registers prompt templates from local files and serves them to clients with the assistant role.
README
mcp-local-prompt-server
An MCP server that dynamically registers prompt templates from local files and serves them to clients with the assistant role (not user).
Features
- File-based prompts — Drop
.txt,.md,.yaml,.yml, or.jsonfiles intoprompts/and they are registered automatically on startup - Clean naming — Filenames like
grammer_correction_agent.mdbecome prompt names like"grammer correction agent" - Assistant role delivery — Prompts are served with
role: "assistant", so clients treat the content as system-level guidance, not user input - Stdio transport — Connects via standard input/output, compatible with any MCP host
Example

The screenshot shows the MCP Inspector listing all 4 prompts after the server starts.
Prerequisites
- Node.js 18+
- npm
Installation
git clone git@github.com:engrtipusultan/mcp-local-prompt-server.git
cd mcp-local-prompt-server
npm install
Included prompts
| File | Prompt name | Description |
|---|---|---|
grammer_correction_agent.md |
grammer correction agent | Fixes English grammar errors with minimal changes, preserving the user's original style |
parapharasing_agent.md |
parapharasing agent | Paraphrases text with configurable tone (professional, casual, etc.) and length adjustments |
prompt_enhancement_agent.md |
prompt enhancement agent | Refines user inputs into structured, actionable prompts |
summerization_agent.md |
summerization agent | Summarizes and structurally maps long texts with metadata, TL;DR, and hierarchical outlines |
Adding more prompts
Create a file in the prompts/ directory:
prompts/
├── code_review_agent.md
├── sql_optimizer.txt
└── writing_feedback.yaml
Supported extensions: .txt, .md, .yaml, .yml, .json. Hyphens and underscores in filenames are converted to spaces for the prompt name. No code changes needed — prompts are registered automatically on the next server start.
Usage
Run
npm run dev
or
npm start
Both run the server directly via tsx — no build step needed.
Testing with MCP Inspector
npx @modelcontextprotocol/inspector npm run dev
This opens a web UI where you can browse registered prompts and invoke them interactively.
How it works
On startup, the server scans the prompts/ directory for supported files. Each file is registered as an MCP prompt using mcpServer.registerPrompt(). When a client requests a prompt, the server reads the file content and returns it as an assistant-role message.
┌─────────────────────────────────────┐
│ MCP Host (client) │
│ stdio transport │
└────────────┬────────────────────────┘
│ JSON-RPC
▼
┌─────────────────────────────────────┐
│ mcp-local-prompt-server │
│ │
│ ┌────────────────────────────────┐ │
│ │ prompts/ │ │
│ │ ├── grammer_correction.md │ │
│ │ ├── summarization_agent.md │ │
│ │ └── ... │ │
│ └────────────────────────────────┘ │
└─────────────────────────────────────┘
Scripts reference
| Script | Command | Description |
|---|---|---|
npm run dev |
tsx server.ts |
Run the server |
npm start |
tsx server.ts |
Run the server (alias) |
License
Apache 2.0
推荐服务器
Baidu Map
百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。
Playwright MCP Server
一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。
Magic Component Platform (MCP)
一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。
Audiense Insights MCP Server
通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。
VeyraX
一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。
graphlit-mcp-server
模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。
Kagi MCP Server
一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。
e2b-mcp-server
使用 MCP 通过 e2b 运行代码。
Neon MCP Server
用于与 Neon 管理 API 和数据库交互的 MCP 服务器
Exa MCP Server
模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。