MCP / Strumenti

Extend IntelligenceBox capabilities with Model Context Protocol (MCP) tools. Connect to databases, APIs, file systems, and more through a standardized plugin system.

What is MCP?

The Model Context Protocol (MCP) is a standard for connecting AI models to external tools and data sources. IntelligenceBox uses MCP to enable assistants to:

Query databases (PostgreSQL, MongoDB)
Access file systems with configurable permissions
Interact with APIs (GitHub, Slack, Stripe)
Perform web searches
Execute custom business logic

Architecture

MCP Architecture

┌─────────────────────────────────────────────────────────────────┐
│                      IntelligenceBox                             │
│                                                                 │
│  ┌─────────────┐         ┌──────────────────┐                  │
│  │  Assistant  │ ──────► │   MCP Client     │                  │
│  │   (LLM)     │         │  (Plugin Mgr)    │                  │
│  └─────────────┘         └────────┬─────────┘                  │
│                                   │                             │
│                    ┌──────────────┼──────────────┐             │
│                    ▼              ▼              ▼             │
│            ┌───────────┐  ┌───────────┐  ┌───────────┐        │
│            │ Plugin A  │  │ Plugin B  │  │ Plugin C  │        │
│            │ (Docker)  │  │ (Docker)  │  │ (Local)   │        │
│            └───────────┘  └───────────┘  └───────────┘        │
│                                                                 │
└─────────────────────────────────────────────────────────────────┘

Key Concepts

Docker Containers

Each MCP plugin runs in an isolated Docker container with its own dependencies and permissions.

Tool Definitions

Plugins expose tools with JSON schemas that describe inputs, outputs, and capabilities.

Secure Configuration

API keys and credentials are configured per-plugin and never exposed to the LLM.

Plugin Lifecycle

Registration: Plugin is added to the registry with its Docker image and config schema
Configuration: User configures the plugin with required credentials/settings
Startup: Docker container is started when an assistant needs the plugin
Tool Discovery: MCP client queries the plugin for available tools
Execution: Assistant invokes tools through the MCP protocol
Shutdown: Container is stopped when no longer needed (configurable)

Communication

MCP supports multiple transport mechanisms:

Transport	Use Case	Notes
`stdio`	Docker containers	Default for most plugins
`http`	Remote services	REST-based communication
`sse`	Streaming responses	Server-Sent Events