Jetski MCP Analytics and Authentication Platform

Jetski is an open-source MCP (Model Context Protocol) Analytics and Authentication Platform, part of the HyprMCP ecosystem. It acts as a management layer for MCP servers by deploying instances of mcp-gateway that sit in front of your existing MCP server.

Key capabilities include:

• Installation & Setup: Automatically generates step-by-step instructions for supported MCP clients to reduce user churn.
• Authentication: Provides zero-code user identification and management via a proxy, allowing for personalized tools and responses.
• Analytics & Visibility: Captures comprehensive prompt analytics (which prompts trigger which tools), real-time debug logs, and usage metrics without requiring changes to your original MCP server code.

Jetski operates as a proxy layer between MCP Clients (like Claude Desktop or ChatGPT) and your MCP Server.

The Request Lifecycle:

1. Client Request: The MCP Client sends a request to the HyprMCP Proxy (the mcp-gateway) instead of directly to your server.
2. Authentication & Validation: The Proxy authenticates the user and validates their permissions.
3. Forwarding: The Proxy forwards the request to your unchanged MCP server.
4. Response & Capture: The Proxy receives the response from your server and automatically captures logs, analytics, and monitoring data.
5. Visibility: All captured data is sent to the Jetski Platform, where it is visualized in real-time dashboards.

If you want to use Kubernetes orchestration for your MCP gateways, use Minikube. This requires the metacontroller to be deployed within your current kubectl context.

1. Start Minikube using the provided mise command.
2. Deploy metacontroller using the serve command with the --install-controller flag.

# Start Minikube
mise run minikube-start

# Deploy metacontroller
mise run serve -- --install-controller

An Organization represents a tenant within the system. It extends a Base interface and contains identity information, creation timestamps, and a settings object which is a combination of domain and authentication settings.

Settings Structure:

• OrganizationDomainSettings: Includes optional customDomain (string).
• OrganizationAuthSettings: Includes authorization which contains dcrPublicClient (boolean).

export interface Organization {
  id: string;
  createdAt: string;
  name: string;
  settings: OrganizationSettings;
}

export type OrganizationSettings = OrganizationDomainSettings & OrganizationAuthSettings;

To apply Shadcn-like styling to a table, use the hlmTable attribute on the <table> element and the corresponding directive attributes on its child elements.

Each child directive (like hlmTHead, hlmTr, etc.) accepts a class input to allow for additional local styling via userClass.

<table hlmTable>
  <thead hlmTHead>
    <tr hlmTr>
      <th hlmTh>Header</th>
    </tr>
  </thead>
  <tbody hlmTBody>
    <tr hlmTr>
      <td hlmTd>Data</td>
    </tr>
  </tbody>
</table>

You can provide a global configuration for all table components using provideHlmTableConfig. This allows you to override default Shadcn-like styles for all table elements (table, thead, tbody, tfoot, tr, th, td, caption) across your application.

Pass a Partial<HlmTableVariant> to the provider. Any properties not specified in your config will fall back to the HlmTableVariantDefault values.

// In your application config or component providers
providers: [
  provideHlmTableConfig({
    table: 'my-custom-table-class',
    tr: 'custom-row-class'
  })
]

The postgres service is used for persistent data storage. It can be configured using the following environment variables:

• POSTGRES_USER: The database user (default: local).
• POSTGRES_PASSWORD: The database password (default: local).
• POSTGRES_DB: The name of the database to create (default: jetski).

Data is persisted in the postgres volume mapped to /var/lib/postgresql/data/.

postgres:
    image: "postgres:17.5-alpine3.20"
    environment:
      POSTGRES_USER: local
      POSTGRES_PASSWORD: local
      POSTGRES_DB: jetski
    volumes:
      - "postgres:/var/lib/postgresql/data/"
    ports:
      - "127.0.0.1:5432:5432"

The application's HttpClient is configured to use withFetch() and includes a custom authInterceptor to handle authentication headers. It also supports interceptors provided via Dependency Injection (DI).

provideHttpClient(
  withInterceptors([authInterceptor]),
  withInterceptorsFromDi(),
  withFetch(),
),

The mailpit service provides an SMTP server and a web UI for inspecting outgoing emails during development.

• SMTP Port: 1025
• Web UI Port: 8025

mailpit:
    image: "axllent/mailpit:v1.27.10"
    ports:
      - "1025:1025"
      - "8025:8025"