Model Context Protocol moderate risk

Access Point

MCP Access Point bridges HTTP services to MCP clients, enabling interaction without server-side changes, using Pingora for high performance and supporting multi-tenancy.

Connections & Capabilities

Connects To

GitHubDocker

Capabilities

readwriteadmin

Quickstart

Install

npx @modelcontextprotocol/inspector

Exposed MCP Tools (5)

safe

GET /admin/resources/{type}/{id}

Retrieves a specific resource of a given type by its ID.

Read-only operation, no modification of system state.

moderate

POST /admin/resources/{type}/{id}

Creates a new resource of a specified type with a given ID.

Creates new resources, potentially affecting system behavior.

high

PUT /admin/resources/{type}/{id}

Updates an existing resource of a specified type with a given ID.

Modifies existing resources, potentially disrupting services.

critical

DELETE /admin/resources/{type}/{id}

Deletes a resource of a specified type with a given ID.

Destructive operation that removes resources from the system.

high

POST /admin/reload/config

Reloads the full configuration from a file, updating the service's behavior.

Can drastically alter system behavior based on the new configuration.

Safety Assessment

The MCP Access Point is relatively safe for read-only operations. However, the Admin API introduces risks due to its ability to modify configurations. Secure deployment requires careful management of the API key and thorough validation of configuration changes.

Supports read-only operations for retrieving data.
Admin API includes optional API key authentication.
Configuration validation available before applying changes.
Supports dynamic configuration updates without service restarts.

Admin API allows modification of upstreams, services, and routes.
Potential for misconfiguration via Admin API leading to service disruption.
Exposes internal HTTP services to MCP clients.
If API key is compromised, unauthorized configuration changes are possible.