Skill סוכני Claude ו-MCP ל-Claude

📖 מה ה-Skill הזה כולל

מתי להשתמש

"Claude Agent", "MCP", "Model Context Protocol", "Anthropic agent", "Claude Code".

הוראות עבודה

1. MCP — Model Context Protocol

Anthropic's open standard for connecting AI assistants to data sources and tools.

Why MCP

Standard interface — write once, use everywhere.
Pre-built servers — many available.
Composable — multiple servers per agent.
Open source ecosystem.

2. Architecture

Claude Agent (Client)
    ↕ (MCP Protocol)
MCP Server 1 (Filesystem)
MCP Server 2 (Database)
MCP Server 3 (API tool)

3. MCP Servers Available 2026

Official (Anthropic)

Filesystem — read/write files.
GitHub — repos, issues.
Google Drive — files.
Slack — messages.
PostgreSQL — DB.
SQLite.
Brave Search.
Puppeteer — browser automation.

Community

100+ MCP servers (github.com/modelcontextprotocol).
Stripe, Notion, Linear, Jira, etc.

4. Where MCP Used

Claude Desktop

Install MCP servers via settings.
Claude has access to those tools.

Claude Code (CLI)

MCP integration native.
Tools available in agent context.

Custom Apps

Build your own client using MCP SDK.

5. Building MCP Server

# Simple MCP server in Python
from mcp.server import Server, Tool

server = Server("my-tools")

@server.tool("get_user")
def get_user(user_id: str) -> dict:
    """Get user by ID."""
    # Your implementation
    return {"id": user_id, "name": "Dana"}

server.run()

Connect to Claude Desktop

// claude_desktop_config.json
{
  "mcpServers": {
    "my-tools": {
      "command": "python",
      "args": ["/path/to/server.py"]
    }
  }
}

6. Claude Agents SDK

Build production agents with Claude as the engine.

Features

Sub-agents — delegate to specialized agents.
Memory — persistent across sessions.
Tool use — MCP-native.
Permission system — control what agent can do.

Use Cases

Coding agents (Claude Code).
Research agents.
Customer support agents.

7. Claude Code (Reference Implementation)

What

CLI agent for coding.
Bash, file edits, MCP tools.
Best example of Claude Agents SDK in action.

Use

Refactoring large codebases.
Implementing features autonomously.
Test creation.
Documentation generation.

8. MCP vs Custom Tool Use

	MCP	Custom Tool Use
Setup	Install server	Build from scratch
Reusability	Cross-app	App-specific
Maintenance	Server author	You
Best for	Common tools	Unique tools

9. Cost

MCP itself: free protocol.
Compute: standard API costs.
Self-hosted MCP servers: hosting cost.

10. Israel Specifics

Hebrew tools can be MCP servers.
Israeli SaaS — build MCP servers for them (HiBob, monday.com, etc.).
Privacy: self-host MCP for sensitive data.

11. Common Pitfalls

❌ Too many MCP servers — context confusion. ❌ No permission scoping — security risk. ❌ Custom tool when MCP exists — wasted effort.

12. Future

MCP standard becoming default for AI agents.
More servers monthly.
Industry adoption growing (OpenAI considering compat).

13. אסיים בהמלצה.

פרומפט לדוגמה

Build MCP server for our internal API.

Use Claude Agents to automate report generation.

MCP vs OpenAI Assistants — choose.

📥 התקנה בחצי דקה

1. הורד ופתח את קובץ ה-ZIP — תקבל תיקייה בשם claude-agents-mcp.

2. ב-Claude Code: העבר את התיקייה אל ~/.claude/skills/.
באפליקציה (Claude / Cowork): הגדרות ← Capabilities ← Skills ← העלאה.

3. בקש מ-Claude את מה שצריך בעברית — הוא יפעיל את ה-skill לבד כשזה רלוונטי.