Skip to main content

Model Context Protocol (MCP)

Arrotech Hub implementation of the Model Context Protocol (MCP) provides a unified interface for AI models to securely interact with external tools and data, without hardcoding platform-specific logic.

The Tool Proxy Architecture​

When you call an MCP tool via Arrotech, we act as a secure proxy:

sequenceDiagram
    participant App as Your AI App
    participant Hub as Arrotech Hub
    participant Registry as Tool Registry
    participant Service as Slack/HubSpot API

    App->>Hub: POST /mcp/call {name: "slack_send", args: {...}}
    Hub->>Registry: Lookup Tool Schema
    Registry-->>Hub: Return Definition
    Hub->>Hub: Validate Args & Inject Secret
    Hub->>Service: Call External API
    Service-->>Hub: API Response
    Hub-->>App: Structured JSON Result

Key Components​

1. Dynamic Discovery​

Use the /mcp/tools endpoint to fetch a list of all tools currently "equipped" in your workspace. This dynamically updates whenever you add a new Connection.

2. JSON Schema Validation​

Every tool has a strictly defined input schema. Arrotech validates your request before it ever reaches the external service, preventing unnecessary API failures and costs.

Example: Sending a Slack Message​

Instead of learning the Slack Web API, you simply use the Arrotech MCP bridge:

{
  "name": "slack_send_message",
  "arguments": {
    "channel": "C12345",
    "text": "Hello from the Arrotech MCP Bridge!"
  }
}

Learn More​