Remote MCP AuthKit Server
Enables remote MCP client connections with WorkOS AuthKit authentication and organization-based permission controls. Demonstrates permission-gated tool access, such as image generation tools restricted by user permissions.
README
Model Context Protocol (MCP) Server + WorkOS AuthKit
This is an example Model Context Protocol (MCP) server that allows remote clients to connect and authenticate using WorkOS AuthKit.
AuthKit supports user management features including an organization-centric authentication model allowing you to control tool access based on user and organization permissions.
Getting Started
First to create a WorkOS account by signing into the WorkOS Dashboard.
Next, add the MCP server's callback URL as a Redirect URI under Redirects
-> Sign in callback. If you are testing locally, this will be
http://localhost:8788/callback, or if deployed, the domain of your deployed
worker with the same /callback path.
Next, you must set the WORKOS_CLIENT_ID and WORKOS_CLIENT_SECRET environment
variables. These can be obtained from the WorkOS Dashboard under API Keys.
You can set these in the Cloudflare dashboard or using the wrangler
CLI:
$ npx wrangler secret put WORKOS_CLIENT_ID <your_workos_client_id>
$ npx wrangler secret put WORKOS_CLIENT_SECRET <your_workos_client_secret>
Note: The WORKOS_CLIENT_ID isn't technically a secret and so you may also choose
to set it via your wrangler.jsonc configuration file. But WORKOS_CLIENT_SECRET is not
public and should be securely set elsewhere.
And that's it! You can now test out your remote MCP server using the example playground below.
Testing MCP Authentication
Visit the Cloudflare Workers AI playground and enter the URL of your worker:
# Local
http://localhost:8788/sse
# Deployed
https://<your-worker-domain>/sse
After clicking Connect, you'll be redirected to your WorkOS AuthKit domain, where you can sign-in, and be returned to the playground authenticated as a WorkOS AuthKit user.
In the demo code, the generateImage tool is gated behind the image_generation permission. You can read more about Permissions in AuthKit here. Try assigning a permission with the image_generation slug to your user to see how it enables additional tools in the playground.
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。