# mcp-inception

> MCP Server

This TypeScript-based MCP server allows delegating tasks and offloading context windows by calling other MCP clients, enabling parallel and map-reduce execution of tasks, and acting as an agent for your agent.

## Overview

- **Category:** AI
- **Language:** JavaScript
- **Stars:** 28
- **Forks:** 1
- **Owner:** tanevanwifferen
- **GitHub:** https://github.com/tanevanwifferen/mcp-inception
- **Created:** 2025-01-25T14:33:16+00:00
- **Updated:** 2025-03-28T22:26:40+00:00
- **Source:** https://model-context-protocol.com/servers/mcp-client-delegation-context-window-offload

## Setup

## Setup

1.  Install mcp-client-cli, including the config file and necessary MCP servers in `~/.llm/config.json`.
2.  Create a bash file to activate the venv and execute the `llm` executable.
3.  Install dependencies using `npm install`.
4.  Build the server using `npm run build`.
5.  For development with auto-rebuild, use `npm run watch`.
6.  To use with Claude Desktop, add the server config to `~/Library/Application Support/Claude/claude_desktop_config.json` (MacOS) or `%APPDATA%/Claude/claude_desktop_config.json` (Windows), specifying the command, args, disabled status, autoApprove list, and environment variables (MCP_INCEPTION_EXECUTABLE and MCP_INCEPTION_WORKING_DIR).
7.  Use the MCP Inspector for debugging with `npm run inspector`.

## Tools

## Available Tools

1.  MCP Server and Client in one (Made with use of [mcp-client-cli](https://github.com/adhikasp/mcp-client-cli))
2.  Offload context windows
3.  Delegate tasks
4.  Parallel and map-reduce execution of tasks
5.  `execute_mcp_client` (Ask a question to a separate LLM, ignore intermediate steps, and return the output.)
6.  `execute_parallel_mcp_client` (Takes a list of inputs and a main prompt, and executes the prompt in parallel for each string in the input.)
7.  `execute_map_reduce_mcp_client` (Process multiple items in parallel and then sequentially reduce the results to a single output.)