Configuration Notice

This implementation requires initial setup; contributions to simplify the configuration process are welcomed via PRs.

MCP Inception Service

This utility enables an executing MCP client to invoke and manage another MCP client instance, effectively creating a nested agent architecture. Key benefits include delegating complex operations and mitigating constraints on the primary context window.

This service is built upon TypeScript and functions as a hybrid MCP server/client.

• Integrates functionality derived from mcp-client-cli.
• Provides mechanisms for context window offloading.
• Facilitates task delegation to subordinate agents.
• Features built-in capabilities for parallel and map-reduce task execution.

Functional Tools

Tool Set

• execute_mcp_client: Sends a direct query to an external LLM system, completely bypassing the reporting of intermediate reasoning steps, and returns only the final answer.
• Parameter: question (Required).
• Output: The final result, stripping all contextual overhead.
• execute_parallel_mcp_client: Manages the concurrent execution of a primary prompt across a collection of input strings. For example, retrieving the current time for six distinct global locations (e.g., London, Paris, Tokyo).
• Parameter: mainPrompt (The instruction to execute for each item).
• Parameter: inputs (An array of strings, one for each parallel execution).
• Operates concurrently on each input item.
• Note: Dependency on this specific PR is required for this functionality.
• execute_map_reduce_mcp_client: Executes a process across multiple items concurrently (Map phase), followed by a sequential consolidation of the results into a singular output (Reduce phase).
• Parameter: mapPrompt (Template using {item} for per-item processing).
• Parameter: reducePrompt (Template using {accumulator} and {result} for aggregation).
• Parameter: items (The collection of data to process).
• Optional Parameter: initialValue (Starting point for the accumulator).
• Workflow: Parallel mapping, followed by sequential reduction.
• Use Case Example: Analyzing several documents and generating a unified summary of key findings.

Implementation Details

Prerequisites

• Install mcp-client-cli.
• Ensure the associated configuration file (~/.llm/config.json) is set up with necessary MCP endpoint definitions.
• Create a shell script to manage environment activation and execution of the llm binary:

#!/bin/bash
source ./venv/bin/activate
llm --no-confirmations

Installation Steps

Install Node package dependencies:

npm install

Build the production server artifact:

npm run build

For active development with automatic recompilation:

npm run watch

Deployment

To integrate this service with the Claude Desktop environment, update your configuration file:

MacOS Location: ~/Library/Application Support/Claude/claude_desktop_config.json Windows Location: %APPDATA%/Claude/claude_desktop_config.json

{
  "mcpServers": {
    "mcp-inception": {
      "command": "node",
      "args": ["~/Documents/Cline/MCP/mcp-inception/build/index.js"], // Path to the compiled output
      "disabled": false,
      "autoApprove": [],
      "env": {
        "MCP_INCEPTION_EXECUTABLE": "./run_llm.sh", // Reference to the startup script above
        "MCP_INCEPTION_WORKING_DIR": "/path/to/mcp-client-cli/workspace"
      }
    }
  }
}

Diagnostic Assistance

Debugging inter-process communication via stdio can be complex. We recommend utilizing the official MCP Inspector, accessible via a package script:

npm run inspector

This command will launch the Inspector, providing a local URL to a browser interface for inspecting communication streams.