Enhanced AutoGen

You run an Enhanced AutoGen MCP Server that enables sophisticated multi‑agent conversations, prompts, resources, and workflows through a standardized Model Context Protocol interface. It lets you create and manage agents, run complex collaboration pipelines, and monitor real-time status and history for your MCP-based applications.

How to use

You interact with the MCP server through an MCP client or integration that can create agents, start workflows, and drive multi‑agent conversations. Start by creating agents that represent roles in your collaboration (for example researchers, reviewers, or code developers), then compose workflows that chain agent actions together. You can execute two‑agent chats for focused exchanges, run group chats for multi‑agent discussions, and use nested or swarm patterns to structure sophisticated problem solving. Enable memory and teachability to let agents remember preferences and past interactions, and use prompts and resources to tailor conversations to your domain.

How to install

Prerequisites you need before installation are Node.js (version 18 or newer) and Python 3.x, plus Git for cloning repositories. Ensure you have an OpenAI API key if you plan to use OpenAI models.

# Prereqs
node -v
python3 -V
git --version

# Install the MCP server via Smithery (automatic client install)
npx -y @smithery/cli install @DynamicEndpoints/autogen_mcp --client claude

# Manual installation steps (clone the repo)
git clone https://github.com/yourusername/autogen-mcp.git
cd autogen-mcp

# Install dependencies
npm install
pip install -r requirements.txt --user

# Build the TypeScript project
npm run build

# Set up configuration
cp .env.example .env
cp config.json.example config.json
# Edit .env and config.json with your settings

Additional configuration and usage notes

Environment variables control core features and credentials. You can customize prompts, resources, workflows, and teachability. The server supports multiple agent types, memory persistence, and quality checks to validate and improve results.

Troubleshooting and tips

If you encounter issues, verify that dependencies are installed, your API key is valid, and the server started correctly. Use the debug mode to collect detailed logs when diagnosing problems.

Examples and quickstart

Here are practical examples you can adapt to your use cases. Create an agent, start a workflow, and initiate a two‑agent chat to validate the setup.

Available tools

create_agent

Create agents with advanced configurations for specialized roles within your MCP environment.

create_workflow

Build and configure complete multi-agent workflows to orchestrate collaborations.

get_agent_status

Retrieve detailed metrics and health information for each agent.

execute_chat

Run an enhanced two-agent conversation to exchange ideas and results.

execute_group_chat

Facilitate multi-agent discussions with coordinated participation.

execute_nested_chat

Create hierarchical conversation structures for layered reasoning.

execute_swarm

Enable swarm-based collaborative problem solving among multiple agents.

execute_workflow

Run predefined workflow templates to automate complex tasks.

manage_agent_memory

Handle learning, memory persistence, and memory optimization for agents.

configure_teachability

Enable or configure agent teachability and knowledge retention.