mcp-management

Multi-server MCP management uses a local `.mcp.json` configuration file to connect, enumerate, and execute tools across configured servers. A script-based CLI reads this configuration to coordinate capabilities and delegate execution to subagents, keeping the main context lean.

To list available tools across MCP servers, navigate to the skill's scripts directory and run `npx ts-node cli.ts list-tools` after configuring `.mcp.json`. The CLI executes multi-server discovery and outputs the available tool definitions for analysis.

Intelligent tool selection in MCP orchestration analyzes tool catalogs to load only relevant capabilities for a specific task. This progressive disclosure mechanism keeps prompts concise by avoiding unnecessary tool definitions and preventing context pollution during execution.

MCP management keeps the main reasoning context clean by delegating tool operations and prompt execution to subagents. This architecture isolates tool definitions and execution payloads, ensuring only essential results return to the primary context.

Yes, managing MCP servers with this skill requires the `@modelcontextprotocol/sdk` dependency. You must run `npm install` in the skill's scripts directory to enable the CLI and execute the multi-server discovery and execution commands.

Yes, the execution engine handles dynamic parameter passing for MCP tools and prompts across multiple configured servers. The CLI reads the `.mcp.json` configuration to route requests and invoke the selected capabilities safely and deterministically.