FutureExcited

---
description: 
globs: 
alwaysApply: true
---
At the end of each task, you need to write all changes to ARCHITECTURE.md.

And by write all changes, I mean modify the architecture of the entire app. If you've changed a function - modify its params/output or whatever you changed.
If you've added a new file - add it to the ARCHITECTURE.md.

FutureExcited

import type { Command, CommandGenerator, CommandOptions } from '../../types.js';
import { loadEnv, loadConfig } from '../../config.js';
import { RunCommand } from '../mcp/run.js';
import { MarketplaceManager } from '../mcp/marketplace.js';
import {
  MCPAuthError,
  MCPConnectionError,
  MCPServerError,
  MCPToolError,
  MCPConfigError,

FutureExcited

### Core Architecture

MCP utilizes a **Host-Client-Server** architecture:

1.  **Host:** This is the primary application integrating with AI/LLMs (e.g., Claude Desktop, an IDE like Cursor or Zed, a custom application). The Host's responsibilities include:
    *   Creating and managing multiple **Client** instances.
    *   Orchestrating interactions between the user, the AI model, and various Clients/Servers.
    *   Enforcing security policies, managing user consent, and handling authorization.
    *   Aggregating context from different sources before presenting it to the LLM.
    *   Handling `sampling` requests initiated by Servers (acting as the intermediary to the actual LLM).