mcp-management

Discover and execute MCP server tools via a subagent workflow.

13|Updated Apr 2, 2025
One-click install
npx skills add https://github.com/kettleofketchup/DraftForge --skill mcp-management-kettleofketchup
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: mcp-management
Source: https://github.com/kettleofketchup/DraftForge/tree/main/.claude/skills/mcp-management
Command: npx skills add https://github.com/kettleofketchup/DraftForge --skill mcp-management-kettleofketchup

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill requires @modelcontextprotocol/sdk, and includes scripts (resource) and references (resource) components.

What problem does it solve?

Claude context management for MCP: This skill enables Claude to discover, analyze, and execute MCP server capabilities without polluting the main context window, enabling safe, multi-server coordination and a persistent catalog of capabilities.

Core Features & Use Cases

  • Intelligent tool/prompt/resource discovery across multiple MCP servers from a single config.
  • Context-efficient operation by delegating discovery and execution to a subagent (mcp-manager).
  • Persistent catalog: stores discovered tools/prompts/resources to assets/tools.json for fast offline reference.
  • Use Case: Orchestrate MCP tasks across several servers, discover capabilities, then execute a chosen tool on a specific server and retrieve results.

Quick Start

  1. Install dependencies for the mcp-management scripts.
  2. Configure MCP servers in .claude/.mcp.json.
  3. Run the CLI to list tools/prompts/resources or call a tool.

Frequently Asked Questions about mcp-management

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I manage multiple MCP servers without polluting the main context window?

You can manage multiple MCP servers without context pollution by using a subagent architecture to delegate tool discovery and execution, keeping the main context window clean while orchestrating tasks across servers.

How do I discover available tools, prompts, and resources across MCP servers?

Discover available tools, prompts, and resources by running a local CLI that reads server configurations from .claude/.mcp.json and queries capabilities via a JSON-RPC interface, persisting results to a catalog for offline reference.

What's the best way to execute a specific tool on a target MCP server via Claude?

The best way to execute a specific tool on a target MCP server is through a local MCP client workflow that orchestrates the task, discovers the chosen tool, and retrieves the execution results via the subagent.

Does MCP management require the @modelcontextprotocol/sdk dependency?

Yes, MCP management requires the @modelcontextprotocol/sdk dependency to establish the local client workflow and JSON-RPC interface needed for discovering and executing server capabilities.

Can I use a persistent catalog for fast offline reference of MCP capabilities?

Yes, you can use a persistent catalog by storing discovered tools, prompts, and resources to assets/tools.json, providing fast offline reference for previously queried MCP server capabilities.

Why does my MCP server context management fail when handling multiple servers?

MCP server context management often fails with multiple servers due to context window pollution; delegating discovery and execution to a subagent architecture prevents this by isolating the capability analysis.