Agent MCP

What rinadelph/Agent-MCP does

The rinadelph/Agent-MCP MCP server coordinates multiple AI agents around a software project. It provides separate tool groups for agent lifecycle management, task orchestration, shared project knowledge, and communication. An MCP client can create specialized agents, assign work, inspect progress, and shut agents down when they are no longer needed.

Project information is stored in a persistent knowledge graph or memory bank. Agents can add architectural decisions and implementation details, inspect stored context, and submit retrieval-augmented questions about the project. This is intended to reduce repeated explanations across sessions and help agents work from common project information.

How it works

The rinadelph/Agent-MCP MCP server can run using the repository's Python implementation or an alternative Node.js/TypeScript implementation. The Python server is started with the agent_mcp.cli module; the Node.js version can be started with the repository's npm scripts or its documented global command.

A connected client calls tools rather than managing agents through a separate application protocol. Typical operations include creating a backend or frontend agent, assigning it a task, checking task dependencies and status, then using direct or broadcast messages to coordinate work. The project also includes a dashboard that can be launched separately for visualizing agents, context entries, and collaboration links.

Setup and configuration

The documented Python setup requires Python 3.10 or newer, Node.js 18 or newer, and npm 9 or newer. After cloning the repository, the README instructs users to create a virtual environment, install dependencies with uv, copy .env.example to .env, and provide an OpenAI API key. The Python server can then be started with uv run -m agent_mcp.cli --port 8080.

The Node.js alternative is located under agent-mcp-node. It uses npm install, environment configuration, and either npm run server or the built npm start command. The dashboard has its own npm dependency installation and development command.

Configuration supports host, port, log level, project directory, and maximum agent count through AGENT_MCP_HOST, AGENT_MCP_PORT, AGENT_MCP_LOG_LEVEL, AGENT_MCP_PROJECT_DIR, and AGENT_MCP_MAX_AGENTS. The CLI also documents an optional authentication token.

Tools and capabilities

The rinadelph/Agent-MCP MCP server exposes these documented tools:

• Create, list, and terminate agents.
• Assign tasks, inspect tasks, and update task status.
• Query project knowledge with ask_project_rag.
• Read and update shared project context.
• Send direct messages, broadcast updates, and request assistance.

It also includes a Python client example that creates an agent, assigns a task, and queries project context.

Limitations and notes

The project is aimed at experienced AI developers who understand MCP, AI coding workflows, and distributed systems concepts. Its documentation notes that the setup and orchestration model may be unsuitable for users seeking a simpler coding assistant. The material does not specify a license or a hosted service, so deployment, maintenance, and OpenAI API costs remain the operator's responsibility.