Logto MCP Server
Enables managing Logto resources (users, applications, organizations, roles) via MCP clients like Cursor or Claude Desktop.
README
Logto MCP Server
A Model Context Protocol server that exposes the Logto Management API to MCP clients (Cursor, Claude Desktop, etc.).
Built with:
@modelcontextprotocol/sdk(stdio transport)@logto/apiManagement SDK
Prerequisites
- A Logto instance (self-hosted OSS or Cloud)
- A Machine-to-machine application in Console → Applications
- An M2M role that includes Logto Management API permissions (
all)
See Interact with Management API for details.
Setup
npm install
npm run build
Environment variables
| Variable | Required | Description |
|---|---|---|
LOGTO_APP_ID / LOGTO_CLIENT_ID |
Yes | M2M application ID |
LOGTO_APP_SECRET / LOGTO_CLIENT_SECRET |
Yes | M2M application secret |
LOGTO_ENDPOINT / LOGTO_BASE_URL |
OSS | Self-hosted Logto base URL |
LOGTO_TENANT_ID |
Cloud | Cloud tenant ID (OSS defaults to default) |
LOGTO_API_INDICATOR |
No | OSS default: https://default.logto.app/api |
Copy .env.example as a reference. For MCP clients, pass variables in the env block (below).
Self-hosted / OSS
LOGTO_ENDPOINT=https://your.logto.endpoint
LOGTO_APP_ID=...
LOGTO_APP_SECRET=...
# Optional: LOGTO_TENANT_ID=default
# Optional: LOGTO_API_INDICATOR=https://default.logto.app/api
Logto Cloud
LOGTO_TENANT_ID=your-tenant-id
LOGTO_APP_ID=...
LOGTO_APP_SECRET=...
Cursor / Claude Desktop / MetaMCP config
Local build:
{
"mcpServers": {
"logto": {
"command": "node",
"args": ["C:/Users/rascal/projects/logto-mcp/build/index.js"],
"env": {
"LOGTO_ENDPOINT": "https://your.logto.endpoint",
"LOGTO_APP_ID": "your-m2m-app-id",
"LOGTO_APP_SECRET": "your-m2m-app-secret"
}
}
}
}
From GitHub repo (no package registry auth):
{
"mcpServers": {
"logto": {
"command": "npx",
"args": ["-y", "github:KDunin/logto-mcp"],
"env": {
"LOGTO_ENDPOINT": "https://your.logto.endpoint",
"LOGTO_APP_ID": "your-m2m-app-id",
"LOGTO_APP_SECRET": "your-m2m-app-secret"
}
}
}
}
From GitHub Packages (after a release is published):
# once in ~/.npmrc (needs a PAT with read:packages)
echo "@kdunin:registry=https://npm.pkg.github.com" >> ~/.npmrc
echo "//npm.pkg.github.com/:_authToken=YOUR_GH_PAT" >> ~/.npmrc
{
"mcpServers": {
"logto": {
"command": "npx",
"args": ["-y", "@kdunin/logto-mcp"],
"env": {
"LOGTO_ENDPOINT": "https://your.logto.endpoint",
"LOGTO_APP_ID": "your-m2m-app-id",
"LOGTO_APP_SECRET": "your-m2m-app-secret"
}
}
}
}
Use an absolute path to build/index.js for local runs. After changing env or rebuilding, restart the MCP client.
Tools
| Tool | Description |
|---|---|
logto_health |
Verify M2M credentials |
list_users / get_user / create_user / update_user / delete_user |
Users |
list_applications / get_application / create_application / update_application / delete_application |
Applications |
list_organizations / get_organization / create_organization / update_organization / delete_organization |
Organizations |
list_organization_members / add_organization_members / remove_organization_member |
Organization members |
list_roles / get_role / create_role / update_role / delete_role |
Roles |
list_user_roles / assign_user_roles |
User ↔ role assignment |
Development
npm run build # compile TypeScript to build/
npm start # run stdio server (expects MCP client on stdin)
npm run dev # tsc --watch
Logging goes to stderr only so stdout stays reserved for MCP JSON-RPC.
Releasing
Releases are automated by .github/workflows/release.yml when you push a version tag.
- Bump
versioninpackage.json(must match the tag without thevprefix). - Commit, then tag and push:
git tag v1.0.1
git push origin v1.0.1
The workflow will:
- Build the package
- Create a GitHub Release with notes and the
.tgzartifact - Publish to GitHub Packages as
@kdunin/logto-mcp(stable tags only; usesGITHUB_TOKEN, no npmjs login)
ENEEDAUTH against registry.npmjs.org means you were targeting the public npm registry. This project publishes to https://npm.pkg.github.com instead.
Prerelease tags like v1.0.1-beta.1 create a GitHub prerelease and skip package publish.
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。