Node Runner MCP Server
Enables running Node.js scripts and npm commands with permission prompts via node-notifier, including server management, script execution with stdin, and npm package documentation fetching.
README
Node Runner MCP Server
An MCP server that allows you to run Node.js scripts and npm commands, with permission prompts via node-notifier.
Requirements
- Node.js >= 22.0.0
Features
- Run Node.js scripts with arguments and standard input
- Execute npm scripts from package.json files with standard input
- Run JavaScript code directly with Node's eval and provide standard input
- Start Node.js servers that continue running in the background
- List running servers and view their status
- Stop running servers gracefully or forcefully when needed
- Retrieve and filter server logs for debugging and monitoring
- Select specific Node.js versions using NVM
- View available npm scripts in package.json files
- Fetch documentation for npm packages with README and metadata
- Permission prompts before any execution (can be disabled)
Setup
- Clone this repository
- Install dependencies:
npm install - Build the TypeScript code:
npm run build
Project Structure
src/index.ts- Main MCP server implementationtest.js- Sample test script to run with the server
Usage with Claude for Desktop
- Build the project with
npm run build - Add the server to your Claude for Desktop configuration:
{
"mcpServers": {
"node-runner": {
"command": "npx",
"args": ["-y", "mcp-node@latest"],
"env": {
"DISABLE_NOTIFICATIONS": "true", // Optional: disable permission prompts
"EVAL_DIRECTORIES": "/path/to/safe/dir1:/path/to/safe/dir2" // Optional: additional allowed eval directories
}
}
}
}
- Restart Claude for Desktop
- You can now ask Claude to run Node.js scripts or npm commands
- (Optional) Use the
envconfiguration to disable notification prompts as shown above
Available Tools
start-node-server
Starts a Node.js server that continues running in the background, even after the command completes.
Parameters:
scriptPath: Path to the Node.js server script to executecwd: Directory to run the server inserverName: (Optional) Friendly name for the server (defaults to the script filename)nodeArgs: (Optional) Arguments to pass to the Node.js executable itselfargs: (Optional) Array of arguments to pass to the server script
Example prompt: "Start an Express server from server.js and keep it running"
Example usage:
start-node-server({
scriptPath: "/absolute/path/to/server.js",
cwd: "/absolute/path/to/project",
serverName: "My Express API",
args: ["--port", "3000"]
});
list-servers
Lists all running Node.js servers started via the MCP server.
Parameters:
showLogs: (Optional) Boolean to include recent logs in the output (default: false)serverId: (Optional) Server ID to get details for a specific server
Example prompt: "Show me all running Node.js servers"
Example to view detailed logs for a specific server:
list-servers({
serverId: "server-1234567890-1234",
showLogs: true
});
stop-server
Stops a running Node.js server.
Parameters:
serverId: ID of the server to stopforce: (Optional) Boolean to force kill the server with SIGKILL instead of SIGTERM (default: false)
Example prompt: "Stop the Node.js server with ID server-1234567890-1234"
Example to forcefully terminate a server:
stop-server({
serverId: "server-1234567890-1234",
force: true
});
get-server-logs
Retrieves the last N lines of logs from a running server with filtering options. This tool is essential for debugging and monitoring server behavior without having to stop it.
Parameters:
serverId: ID of the server to get logs fromlines: (Optional) Number of log lines to retrieve (default: 50)filter: (Optional) String to filter logs (case-insensitive)stdout: (Optional) Boolean to include stdout logs (default: true)stderr: (Optional) Boolean to include stderr logs (default: true)
Key features:
- Retrieves logs from both running and exited servers
- Filters logs to show only stdout or stderr as needed
- Searches for specific text within logs
- Shows server status alongside the logs
- Limits output to exactly the number of lines requested
Example prompt: "Show me the last 100 logs from the server with ID server-1234567890-1234"
Example to view only error output:
get-server-logs({
serverId: "server-1234567890-1234",
stderr: true,
stdout: false
});
Example with text filtering:
get-server-logs({
serverId: "server-1234567890-1234",
lines: 100,
filter: "error"
});
run-node-script
Executes a Node.js script file.
Parameters:
scriptPath: Path to the Node.js script to executenodeArgs: (Optional) Arguments to pass to the Node.js executable itselfargs: (Optional) Array of arguments to pass to the scriptstdin: (Optional) Text to provide as standard input to the scriptcwd: (Optional) Directory to run the script in (defaults to OS temp directory if not specified)timeout: (Optional) Timeout in milliseconds after which the process is killed (defaults to 60000ms)
Example prompt: "Run the test.js script with arguments 'hello' and 'world'"
Example with working directory:
run-node-script({
scriptPath: "/absolute/path/to/my-script.js",
args: ["arg1", "arg2"],
cwd: "/absolute/path/to/project"
});
Example with timeout:
run-node-script({
scriptPath: "/absolute/path/to/long-running-script.js",
timeout: 10000 // Kill after 10 seconds
});
run-npm-script
Executes an npm script from a package.json file.
Parameters:
packageDir: Directory containing the package.jsonscriptName: Name of the script to runargs: (Optional) Array of arguments to pass to the scriptstdin: (Optional) Text to provide as standard input to the script
Example prompt: "Run the 'start' script from the package.json in the current directory"
run-npm-install
Executes npm install to install all dependencies or a specific package.
Parameters:
packageDir: Directory containing package.jsondependency: (Optional) Specific dependency to install (leave empty to install all dependencies from package.json)
Example prompt: "Install express in the current project"
Example usage:
run-npm-install({
packageDir: "/absolute/path/to/project",
dependency: "express"
});
run-node-eval
Executes JavaScript code directly.
Parameters:
code: JavaScript code to executeevalDirectory: (Optional) Directory to execute the code instdin: (Optional) Text to provide as standard input to the codetimeout: (Optional) Timeout in milliseconds after which the process is killed (defaults to 5000ms)
Example prompt: "Run this JavaScript code: console.log('Hello world');"
Example with timeout:
run-node-eval({
code: `
// This code would be terminated after 3 seconds
console.log('Starting long operation...');
setTimeout(() => {
console.log('This will never be logged');
}, 5000);
`,
timeout: 3000 // Kill after 3 seconds
});
fetch-npm-docs
Fetches documentation for an npm module, including README and metadata. Downloads the package, extracts the README, and caches results to avoid redundant downloads.
Parameters:
packageName: Name of the npm packageversion: (Optional) Specific version to fetch (defaults to latest)
Key features:
- Retrieves package metadata using
npm view - Downloads and extracts the package tarball to get the README
- Caches packages both in memory and on disk to avoid redundant downloads
- Works with the selected Node.js version
Example prompt: "Show me the documentation for the express package"
Example usage:
fetch-npm-docs({
packageName: "express",
version: "4.18.2"
});
list-node-versions
Lists all available Node.js versions installed via NVM (Node Version Manager).
Parameters: None
Example prompt: "Show me all installed Node.js versions"
select-node-version
Selects a specific Node.js version to use for subsequent script executions.
Parameters:
version: Node.js version to use (e.g., 'v18.20.5', 'system', 'lts/*', or other NVM aliases)
Example prompt: "Use Node.js version 18 for running scripts"
get-node-version
Displays information about the currently selected Node.js version.
Parameters: None
Example prompt: "What Node.js version is currently being used?"
Examples of Using Standard Input
Passing Input to a Node Script
// Example script: process-data.js
process.stdin.on('data', (data) => {
const input = data.toString().trim();
console.log(`Received: ${input}`);
// Process the input...
});
You can execute this with standard input and a specific working directory:
run-node-script({
scriptPath: "/absolute/path/to/process-data.js",
stdin: "This is input data",
cwd: "/absolute/path/to/my-project-directory" // Sets the working directory for the script
});
Using Standard Input with Eval
run-node-eval({
code: `
let data = '';
process.stdin.on('data', (chunk) => { data += chunk; });
process.stdin.on('end', () => {
console.log('Received:', data);
});
`,
stdin: "Data to process"
});
Reading a File Then Using It As Standard Input
// First read the file
const fileContent = read_file({ path: "/absolute/path/to/data.txt" });
// Then pass it as standard input to a script
run-node-script({
scriptPath: "/absolute/path/to/process-data.js",
stdin: fileContent,
cwd: "/absolute/path/to/working-directory"
});
Server Management and Monitoring
This MCP server provides a complete solution for running and monitoring Node.js servers in the background:
-
Starting Servers: Use
start-node-serverto launch servers that keep running even after the command completes. -
Monitoring: Monitor server logs in real-time with
get-server-logsto keep track of activity and troubleshoot issues. -
Process Management: View all running servers with
list-serversand get detailed information about their status. -
Graceful Shutdown: Stop servers gracefully with
stop-server, preserving any in-flight operations.
Example Server Monitoring Workflow:
// 1. Start a server
const serverInfo = start-node-server({
scriptPath: "/path/to/server.js",
cwd: "/path/to/project",
serverName: "API Server"
});
// Extract server ID from the response
const serverId = serverInfo.content[0].text.match(/Server ID: ([\w-]+)/)[1];
// 2. Monitor logs in real-time (periodic polling)
get-server-logs({
serverId: serverId,
lines: 20
});
// 3. Filter logs for errors only
get-server-logs({
serverId: serverId,
filter: "error",
lines: 50
});
// 4. When finished, stop the server
stop-server({
serverId: serverId
});
Debugging Complex Issues:
When troubleshooting server issues, you can use a combination of tools:
- Check server status with
list-servers - View filtered logs with
get-server-logs - If the server is unresponsive, force stop it with
stop-server({ serverId, force: true })
Resources
node-version
Displays information about the Node.js environment running the MCP server.
URI template: node-version://info
Example prompt: "What version of Node.js is being used to run the scripts?"
npm-scripts
Lists all available npm scripts in a package.json file.
URI template: npm-scripts://{directory}
Example prompt: "Show me the available npm scripts in this project"
Security Considerations
- The server will always prompt for permission before executing any command
- Scripts run with the same permissions as the MCP server process
- Be cautious when running scripts from untrusted sources
Environment Variables
DISABLE_NOTIFICATIONS
Set DISABLE_NOTIFICATIONS=true to automatically approve all permission requests without showing notification prompts:
# Run with notifications disabled
DISABLE_NOTIFICATIONS=true npm run dev
This is useful for automation scenarios or when you don't want to be prompted for each action.
EVAL_DIRECTORIES
Specify a colon-separated list of directories where JavaScript code can be evaluated using the run-node-eval tool:
# Allow code evaluation in specific directories
EVAL_DIRECTORIES=/path/to/dir1:/path/to/dir2 npm run dev
By default, only the system temporary directory is allowed. This environment variable lets you add additional safe directories.
License
MIT
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。