@verdaccio/mcp-server
An MCP server that connects AI clients to a Verdaccio private npm registry, enabling package search, publishing, and registry management via natural language.
README
@verdaccio/mcp-server
⚠️ Experimental: This project is a work in progress and subject to change. APIs, tool names, and behaviour may change without notice between versions.
An MCP (Model Context Protocol) server that connects any MCP-compatible AI client to a Verdaccio private npm registry.
MCP is an open protocol — this server works with any MCP client, not just Claude.
API endpoint paths are sourced directly from @verdaccio/middleware constants (e.g. SEARCH_API_ENDPOINTS, PACKAGE_API_ENDPOINTS) so they stay in sync with Verdaccio's own routing definitions.
Compatible clients
| Client | Config location |
|---|---|
| Claude Desktop / Claude Code | ~/.claude.json or workspace .mcp.json |
| Cursor | .cursor/mcp.json |
| Windsurf | ~/.codeium/windsurf/mcp_config.json |
| VS Code + Copilot | .vscode/mcp.json |
| Continue.dev | ~/.continue/config.json |
| Any custom app | MCP SDK |
Installation
npm install -g @verdaccio/mcp-server
# or run without installing
npx @verdaccio/mcp-server
Using with Claude
Claude Code (CLI)
The quickest way is via the claude mcp add command:
# Local scope (current project only)
claude mcp add verdaccio -e VERDACCIO_URL=http://your-verdaccio:4873 -- npx @verdaccio/mcp-server
# User scope (all projects)
claude mcp add -s user verdaccio -e VERDACCIO_URL=http://your-verdaccio:4873 -- npx @verdaccio/mcp-server
# With auth token
claude mcp add verdaccio \
-e VERDACCIO_URL=http://your-verdaccio:4873 \
-e VERDACCIO_TOKEN=your-token \
-- npx @verdaccio/mcp-server
Claude Desktop
Add to ~/.claude.json:
{
"mcpServers": {
"verdaccio": {
"command": "npx",
"args": ["@verdaccio/mcp-server"],
"env": {
"VERDACCIO_URL": "http://your-verdaccio:4873",
"VERDACCIO_TOKEN": "your-token"
}
}
}
}
Restart Claude after adding the config. You can verify the server is connected by asking:
"What tools do you have available for Verdaccio?"
Tools
start_registry
Start the Verdaccio registry and return its URL. When no VERDACCIO_URL is configured, an embedded registry is started automatically on a random local port with temporary storage.
No parameters required.
Example prompt: "Start a local Verdaccio registry so I can publish packages"
find_package
Search for packages in the registry by name or keyword.
| Parameter | Type | Required | Description |
|---|---|---|---|
query |
string | yes | Package name or search term |
size |
number | no | Max results to return (1–100, default 20) |
Uses endpoint: SEARCH_API_ENDPOINTS.search → /-/v1/search
Example prompt: "Find all packages related to authentication in our private registry"
get_package
Get detailed metadata for a specific package — description, author, license, dist-tags, and version history.
| Parameter | Type | Required | Description |
|---|---|---|---|
package |
string | yes | Package name, e.g. my-lib or @scope/my-lib |
version |
string | no | Specific version e.g. 1.2.3 (omit for latest) |
Uses endpoint: PACKAGE_API_ENDPOINTS.get_package_by_version → /:package{/:version}
Example prompt: "What versions of @myorg/ui-components are available?"
publish_package
Publish a package to the registry by name and version. Uses generatePackageMetadata from @verdaccio/test-helper to generate the full publish payload automatically — no tarball needed.
| Parameter | Type | Required | Description |
|---|---|---|---|
name |
string | yes | Package name, e.g. my-lib or @scope/my-lib |
version |
string | yes | Version to publish, e.g. 1.0.0 |
tag |
string | no | Dist-tag to apply (default: latest) |
Uses endpoint: PUBLISH_API_ENDPOINTS.add_package → /:package
Example prompt: "Publish @myorg/ui-components version 2.1.0 to our private registry"
login
Authenticate with username and password. Returns a Bearer token.
| Parameter | Type | Required | Description |
|---|---|---|---|
username |
string | yes | Verdaccio username |
password |
string | yes | Verdaccio password |
Uses endpoint: USER_API_ENDPOINTS.add_user → /-/user/:org_couchdb_user
get_profile
Get the profile of the currently authenticated user (requires VERDACCIO_TOKEN).
Uses endpoint: PROFILE_API_ENDPOINTS.get_profile → /-/npm/v1/user
list_tokens
List all API tokens for the authenticated user.
Uses endpoint: TOKEN_API_ENDPOINTS.get_tokens → /-/npm/v1/tokens
create_token
Create a new API token for the authenticated user.
| Parameter | Type | Required | Description |
|---|---|---|---|
password |
string | yes | Current user password |
readonly |
boolean | no | Make token read-only (default false) |
cidr |
string[] | no | IP ranges to whitelist, e.g. ["10.0.0.0/8"] |
Uses endpoint: TOKEN_API_ENDPOINTS.get_tokens → /-/npm/v1/tokens (POST)
delete_token
Delete an API token by its key.
| Parameter | Type | Required | Description |
|---|---|---|---|
tokenKey |
string | yes | Token key identifier (from list_tokens) |
Uses endpoint: TOKEN_API_ENDPOINTS.delete_token → /-/npm/v1/tokens/token/:tokenKey
Example

Setup
1. Install and build
npm install
npm run build
2. Configure environment
| Variable | Default | Description |
|---|---|---|
VERDACCIO_URL |
http://localhost:4873 |
Base URL of your Verdaccio instance |
VERDACCIO_TOKEN |
— | Bearer token for authenticated registries |
3. Add to your MCP client
The server config format is the same across all clients — only the config file location differs (see Compatible clients above).
{
"mcpServers": {
"verdaccio": {
"command": "npx",
"args": ["@verdaccio/mcp-server"],
"env": {
"VERDACCIO_URL": "http://your-verdaccio:4873",
"VERDACCIO_TOKEN": "your-token"
}
}
}
}
Development
Testing locally without publishing
To connect your MCP client to the local source without publishing to npm, point it directly at the local file.
Option A — compiled (requires npm run build after each change):
{
"mcpServers": {
"verdaccio": {
"command": "node",
"args": ["/path/to/mcp-server-verdaccio/dist/index.js"],
"env": {
"VERDACCIO_URL": "http://localhost:4873"
}
}
}
}
Option B — TypeScript directly via tsx (no build step, recommended for dev):
{
"mcpServers": {
"verdaccio": {
"command": "npx",
"args": ["tsx", "/path/to/mcp-server-verdaccio/src/index.ts"],
"env": {
"VERDACCIO_URL": "http://localhost:4873"
}
}
}
}
With option B, just save your changes and restart the MCP client — no build step needed.
Running standalone
# Run directly with tsx (no build step)
VERDACCIO_URL=http://localhost:4873 npm run dev
# Build and run compiled output
npm run build
npm start
# Run tests
npm test
npm run test:watch
Stack
@modelcontextprotocol/sdk— MCP server framework@verdaccio/middleware— Verdaccio API endpoint constantszod— Tool input schema validation
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。