npm-mcp
A read-only MCP server for the public npm registry that enables AI clients to search for npm packages and retrieve focused package metadata, versions, and latest dist-tags without exposing install or publish capabilities.
README
npm-mcp
npm-mcp is a small, read-only Model Context Protocol server for the public npm registry. It lets MCP-compatible AI clients discover npm packages and inspect focused package metadata without exposing install, publish, login, shell, filesystem, or arbitrary-fetch capabilities.
Start with Docker
See the Docker quick start to launch the server and MCP Inspector from a fresh checkout. The essential command is:
docker compose --profile inspector up --build -d
To test model-driven MCP tool use without running a model locally, use LibreChat with OpenRouter's free-model router:
docker compose -f compose.yaml -f compose.openrouter.yaml --profile openrouter up --build -d
Then open http://127.0.0.1:3080. An OpenRouter API key is required even for free models. See OPENROUTER_QUICKSTART.md for setup and usage. The local Ollama profile remains available as an optional fully local alternative.
Tools
| Tool | Use it when | Input |
|---|---|---|
npm_search_packages |
The exact package name is unknown or alternatives are wanted | query, optional limit (default 10, max 50) |
npm_get_package |
The exact package name is known and metadata or dependencies are needed | packageName |
npm_get_latest_version |
Only the current latest dist-tag is needed |
packageName |
npm_get_versions |
Published versions and all dist-tags are needed | packageName |
All tools return JSON as text for broad client compatibility and expose the same value as MCP structured content. Package responses are deliberately filtered rather than returning the registry's full metadata document.
Requirements and installation
- Node.js 20 or newer (for native
fetchandAbortSignal.timeout) - npm
git clone <repository-url>
cd npm-mcp
npm install
Run locally
npm start
The process waits for MCP messages on stdin and writes protocol responses to stdout. Do not type interactively into it; an MCP client should launch it. Development watch mode is available with npm run dev.
Deployment
For local clients, use the stdio entry point. For Docker and cloud deployment, use the stateless Streamable HTTP entry point at /mcp; it requires a bearer token by default and exposes /healthz for health probes. See the Docker and cloud deployment guide for Docker Compose and deployment instructions for Google Cloud Run, Azure Container Apps, AWS App Runner/ECS Fargate, and Kubernetes.
The Compose file also includes the official MCP Inspector as an optional client. Run docker compose --profile inspector up --build -d, then open http://127.0.0.1:6274. See Run an MCP client in Docker for token setup and CLI usage.
Test
npm test
Tests use Node's built-in test runner and mocked fetch responses, so they do not require the live npm registry.
To inspect the server interactively with MCP Inspector:
npx @modelcontextprotocol/inspector node src/index.js
Open the URL printed by Inspector, connect, select Tools, choose a tool, enter its arguments, and run it.
MCP client configuration
Use an absolute path so the client can start the server from any working directory:
{
"mcpServers": {
"npm": {
"command": "node",
"args": ["/absolute/path/to/npm-mcp/src/index.js"]
}
}
}
The exact outer configuration key can vary by MCP client, but the command and arguments are the same.
Example calls
{"name":"npm_search_packages","arguments":{"query":"postgres orm","limit":10}}
{"name":"npm_get_package","arguments":{"packageName":"@nestjs/core"}}
{"name":"npm_get_latest_version","arguments":{"packageName":"react"}}
{"name":"npm_get_versions","arguments":{"packageName":"axios"}}
Architecture
src/index.jsowns stdio startup only.src/http.jsandsrc/http-server.jsprovide the authenticated Streamable HTTP entry point used by containers.src/server.jscreates the MCP server and registers the four tools.src/tools/contains schemas, focused response shaping, and handlers.src/services/npm-registry.jsis the only module that communicates with npm. It validates package names, constrains URLs to the official registry endpoints, applies a timeout, and normalizes network/HTTP/response errors.
Scoped names are encoded as a single registry path segment. The server is stateless and performs no writes.
Future improvements
Possible later additions include download statistics, dependency analysis, package security information, caching, and OAuth-based remote authentication.
Recommended Servers
playwright-mcp
A Model Context Protocol server that enables LLMs to interact with web pages through structured accessibility snapshots without requiring vision models or screenshots.
Magic Component Platform (MCP)
An AI-powered tool that generates modern UI components from natural language descriptions, integrating with popular IDEs to streamline UI development workflow.
Audiense Insights MCP Server
Enables interaction with Audiense Insights accounts via the Model Context Protocol, facilitating the extraction and analysis of marketing insights and audience data including demographics, behavior, and influencer engagement.
VeyraX MCP
Single MCP tool to connect all your favorite tools: Gmail, Calendar and 40 more.
graphlit-mcp-server
The Model Context Protocol (MCP) Server enables integration between MCP clients and the Graphlit service. Ingest anything from Slack to Gmail to podcast feeds, in addition to web crawling, into a Graphlit project - and then retrieve relevant contents from the MCP client.
Kagi MCP Server
An MCP server that integrates Kagi search capabilities with Claude AI, enabling Claude to perform real-time web searches when answering questions that require up-to-date information.
Neon Database
MCP server for interacting with Neon Management API and databases
E2B
Using MCP to run code via e2b.
Exa Search
A Model Context Protocol (MCP) server lets AI assistants like Claude use the Exa AI Search API for web searches. This setup allows AI models to get real-time web information in a safe and controlled way.
Qdrant Server
This repository is an example of how to create a MCP server for Qdrant, a vector search engine.