Agent

Agent-MCP enables coordinated multi-agent AI development by tying together specialized agents, a persistent knowledge graph, and real-time task orchestration through MCP compatible clients. It helps you scale complex projects, keep context fresh, and visualize how your agents collaborate to deliver reliable software faster.

How to use

Use an MCP client to connect to the Agent-MCP server and start orchestrating a team of specialized agents. You can spawn workers, query the shared memory graph, assign tasks, and monitor progress in real time. The system surfaces a dashboard to track agent activity, task status, and knowledge graph updates so you always know what your team is doing and why.

How to install

Prerequisites you need to install and run Agent-MCP are a recent Python and Node.js toolchain, plus an OpenAI API key for embeddings and RAG.

Step by step commands to set up and run locally:

# Prerequisites check (ensure versions meet minimums)
python --version  # Should be >=3.10
node --version    # Should be >=18.0.0
npm --version     # Should be >=9.0.0

# Optional: use Node Version Manager if you manage Node with NVM
nvm use

# Quick Python-based setup (recommended)
git clone https://github.com/rinadelph/Agent-MCP.git
cd Agent-MCP
cp .env.example .env  # Add your OpenAI API key
uv venv
uv install
uv run -m agent_mcp.cli --port 8080 --project-dir path-to-directory

# Quick Node.js/TypeScript setup (alternative)
git clone https://github.com/rinadelph/Agent-MCP.git
cd Agent-MCP/agent-mcp-node
npm install
cp .env.example .env  # Add your OpenAI API key
npm run server

"}]} ,{

## Additional sections

MCP server configuration and runtime details are provided to let you connect MCP clients, construct your server endpoints, and manage agents securely. You can run the MCP server over HTTP for remote clients or use a local stdio command for development workflows.

Key MCP setup snippets are shown below so you can connect clients like Claude Desktop or other MCP-enabled tools and begin orchestrating your AI team.

MCP configuration and environment variables are defined in the provided examples. For a server you can run locally, you’ll typically start the MCP endpoint and then point clients to the corresponding HTTP URL or WebSocket endpoint.

## Available tools

### create\_agent

Spawn specialized agents (backend, frontend, testing, etc.)

### list\_agents

View all active agents and their status

### terminate\_agent

Safely shut down agents

### assign\_task

Delegate work to specific agents

### view\_tasks

Monitor task progress and dependencies

### update\_task\_status

Track completion and blockers

### ask\_project\_rag

Query the persistent knowledge graph

### update\_project\_context

Add architectural decisions and patterns

### view\_project\_context

Access stored project information

### send\_agent\_message

Direct messaging between agents

### broadcast\_message

Send updates to all agents

### request\_assistance

Escalate complex issues