MCP Agent Orchestrator

Example Use Cases

• 1

  Coordinate multi-step workflows and transition between planning, researching, executing, and review states.

• 2

  Persist and retrieve conversation context and knowledge across state transitions for consistent long-running tasks.

• 3

  Expose and call structured tools, prompts, and resources through the MCP standard (e.g., when integrating with Claude for Desktop or other MCP-capable clients).

Description

This repository provides a Python implementation of a state-machine-based orchestration server using the Model Context Protocol (MCP). It exposes resources, tools, and prompts to LLMs and supports states like IDLE, PLANNING, RESEARCHING, EXECUTING, REVIEWING, and ERROR. The project includes an example server (orchestrator.py) and a client (orchestrator_client.py) for testing and can be integrated with Claude for Desktop or used with the MCP CLI. It is intended as a customizable template for building agent orchestration and multi-step task workflows.