Clerk MCP Server
Enables AI assistants to query and manage Clerk organizations, users, members, and metadata via the Model Context Protocol. It supports both hosted and per-request secret key modes for flexible integration with various AI environments.
README
Clerk MCP Server
A Model Context Protocol (MCP) server for Clerk
Query and manage your Clerk organizations, members, users, roles, and metadata directly from AI assistants like Claude, Cursor, VS Code Copilot, Windsurf, and more.
Getting Your Clerk API Key
You need a Clerk Secret Key to use this server.
- Go to dashboard.clerk.com and sign in (or create an account)
- Select your application (or create one)
- Navigate to Configure → API Keys
- Copy the Secret Key — it starts with
sk_test_(development) orsk_live_(production)
Never commit your secret key to git. Use environment variables or pass it via headers.
Two Operating Modes
The server supports two modes, auto-detected at startup:
Hosted Mode (Private)
The server owns the Clerk secret key. Set CLERK_SECRET_KEY in your .env file and all requests use that single key. No headers needed from clients.
Best for: Personal use, internal teams, self-hosted deployments.
Public Mode (Per-Request Key)
No secret key on the server. Each client passes their own key via the X-Clerk-Secret-Key HTTP header on every request. The server creates a fresh Clerk client per request.
Best for: Shared deployments, multi-tenant setups, or when you don't want the key stored on the server.
| Hosted Mode | Public Mode | |
|---|---|---|
| Key stored on server | Yes (in .env) |
No |
| Key sent per request | No | Yes (via header) |
| Setup complexity | Simpler | Slightly more config |
| Multi-user support | Single Clerk account | Multiple Clerk accounts |
Quick Start
1. Clone & install
git clone https://github.com/BalajiSriraman/Clerk-MCP.git
cd Clerk-MCP
npm install
2. Configure & run
Hosted mode — set the key once on the server:
cp .env.example .env
# Edit .env and paste your secret key:
# CLERK_SECRET_KEY=sk_test_xxxxxxxxxxxxxxxxxxxxxxxxxxxxx
npm run dev
Public mode — no .env needed, clients provide the key:
npm run dev
Your MCP endpoint is now live at https://clerk-mcp.vercel.app/mcp.
3. Verify
curl https://clerk-mcp.vercel.app/api/health
- Hosted mode:
{"status":"ok","mode":"hosted","clerkConnected":true} - Public mode:
{"status":"ok","mode":"public","clerkConnected":null}
Connect to AI Assistants
Claude Code (CLI)
Hosted mode:
claude mcp add --transport http clerk https://clerk-mcp.vercel.app/mcp
Public mode:
claude mcp add --transport http \
--header "X-Clerk-Secret-Key: sk_test_xxxxxxxxxxxxxxxxxxxxxxxxxxxxx" \
clerk https://clerk-mcp.vercel.app/mcp
Claude Desktop
Add to ~/Library/Application Support/Claude/claude_desktop_config.json (macOS) or %APPDATA%\Claude\claude_desktop_config.json (Windows):
Hosted mode:
{
"mcpServers": {
"clerk": {
"command": "npx",
"args": ["mcp-remote", "https://clerk-mcp.vercel.app/mcp"]
}
}
}
Public mode:
{
"mcpServers": {
"clerk": {
"command": "npx",
"args": [
"mcp-remote",
"https://clerk-mcp.vercel.app/mcp",
"--header",
"X-Clerk-Secret-Key: sk_test_xxxxxxxxxxxxxxxxxxxxxxxxxxxxx"
]
}
}
}
Cursor
Go to Settings → MCP → Add Server:
Hosted mode:
{
"mcpServers": {
"clerk": {
"url": "https://clerk-mcp.vercel.app/mcp"
}
}
}
Public mode:
{
"mcpServers": {
"clerk": {
"url": "https://clerk-mcp.vercel.app/mcp",
"headers": {
"X-Clerk-Secret-Key": "sk_test_xxxxxxxxxxxxxxxxxxxxxxxxxxxxx"
}
}
}
}
VS Code (GitHub Copilot)
Add to your workspace .vscode/mcp.json:
Hosted mode:
{
"mcp": {
"servers": {
"clerk": {
"url": "https://clerk-mcp.vercel.app/mcp"
}
}
}
}
Public mode:
{
"mcp": {
"servers": {
"clerk": {
"url": "https://clerk-mcp.vercel.app/mcp",
"headers": {
"X-Clerk-Secret-Key": "sk_test_xxxxxxxxxxxxxxxxxxxxxxxxxxxxx"
}
}
}
}
}
Windsurf
Add to MCP settings:
Hosted mode:
{
"context_servers": {
"clerk": {
"source": "custom",
"command": "npx",
"args": ["mcp-remote", "https://clerk-mcp.vercel.app/mcp"],
"env": {}
}
}
}
Public mode:
{
"context_servers": {
"clerk": {
"source": "custom",
"command": "npx",
"args": [
"mcp-remote",
"https://clerk-mcp.vercel.app/mcp",
"--header",
"X-Clerk-Secret-Key: sk_test_xxxxxxxxxxxxxxxxxxxxxxxxxxxxx"
],
"env": {}
}
}
}
Docker
Build the image
docker build -t clerk-mcp .
Run in hosted mode
docker run -d -p 3000:3000 -e CLERK_SECRET_KEY=sk_test_xxxxxxxxxxxxxxxxxxxxxxxxxxxxx clerk-mcp
Run in public mode
docker run -d -p 3000:3000 clerk-mcp
Docker Compose
Both modes are defined in docker-compose.yml:
# Hosted mode on port 3000 — set CLERK_SECRET_KEY in .env first
docker compose up clerk-mcp-hosted
# Public mode on port 3001 — no key needed on the server
docker compose up clerk-mcp-public
Available Tools
Organizations
| Tool | Description |
|---|---|
clerk_list_organizations |
List orgs with filtering by name/slug, pagination, and member counts |
clerk_get_organization |
Get org details by ID or slug (includes metadata, timestamps) |
clerk_create_organization |
Create a new organization with name, slug, and metadata |
clerk_update_organization_metadata |
Update org public/private metadata |
clerk_delete_organization |
Delete an organization permanently (irreversible) |
Members
| Tool | Description |
|---|---|
clerk_list_organization_members |
List org members with roles, user data, and metadata |
clerk_update_member_role |
Change a member's role (e.g. org:admin, org:member) |
clerk_update_member_metadata |
Update membership public/private metadata |
clerk_remove_member |
Remove a member from an organization |
Invitations
| Tool | Description |
|---|---|
clerk_list_organization_invitations |
List invitations by status (pending/accepted/revoked) |
clerk_create_invitation |
Invite a user to an org by email |
Users
| Tool | Description |
|---|---|
clerk_list_users |
List all instance users with search by name/email/phone |
clerk_get_user |
Get full user profile: emails, phones, external accounts, metadata |
clerk_update_user_metadata |
Update user public/private/unsafe metadata |
Example Prompts
Once connected, try asking your AI assistant:
List all organizations in my Clerk instance.
How many members does the "engineering" organization have?
Create a new organization called "Design Team" with slug "design-team".
Show me all pending invitations for organization org_2abc123.
Find all users with email addresses containing "@example.com".
Update the public metadata for user user_2xyz789 to set plan: "pro".
Deploying to Production
Build and deploy as any Nuxt/Node.js app:
npm run build
node .output/server/index.mjs
Or use Docker:
docker build -t clerk-mcp .
docker run -p 3000:3000 -e CLERK_SECRET_KEY=sk_live_xxx clerk-mcp
If you self-host, use your own deployment URL instead of https://clerk-mcp.vercel.app in the client configs above.
Compatible with: Vercel, Netlify, Cloudflare Workers (with nitro preset), Railway, Fly.io, or any Node.js host.
Adding New Tools
Create a new file in server/mcp/tools/ — it's automatically discovered by the MCP toolkit:
// server/mcp/tools/clerk-my-new-tool.ts
import { z } from "zod";
export default defineMcpTool({
name: "clerk_my_new_tool",
description: "Description of what this tool does",
inputSchema: {
param: z.string().describe("Parameter description"),
},
annotations: {
readOnlyHint: true,
destructiveHint: false,
openWorldHint: true,
},
async handler({ param }) {
const clerk = useClerkClient();
const result = await clerkCall(() => clerk.someApi.someMethod({ param }));
return jsonResult(result);
},
});
Tech Stack
- Nuxt 4 — Full-stack Vue framework
- @nuxtjs/mcp-toolkit — MCP server module for Nuxt
- @clerk/backend — Clerk Backend SDK
- Zod — Schema validation
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。