Agent-MCP
Author: rinadelph
Description: Multi-Agent Collaboration Protocol (MCP) framework for coordinated AI software development. Provides an MCP server that exposes multi-agent orchestration (agent lifecycle, task assignment, shared knowledge graph/RAG, messaging) plus an optional real-time web dashboard for visualization.
Stars: 1.2k
Forks: 151
License: GNU AGPL-3.0
Category: Open Source
Overview
Installation
# Clone and setup
git clone https://github.com/rinadelph/Agent-MCP.git
cd Agent-MCP
# Check version requirements
python --version # Should be >=3.10
node --version # Should be >=18.0.0
npm --version # Should be >=9.0.0
# If using nvm for Node.js version management
nvm use # Uses the version specified in .nvmrc
# Configure environment
cp .env.example .env # Add your OpenAI API key
uv venv
uv install
# Start the server
uv run -m agent_mcp.cli --port 8080 --project-dir path-to-directory
# Launch dashboard (recommended for full experience)
cd agent_mcp/dashboard && npm install && npm run dev# Clone and setup
git clone https://github.com/rinadelph/Agent-MCP.git
cd Agent-MCP/agent-mcp-node
# Install dependencies
npm install
# Configure environment
cp .env.example .env # Add your OpenAI API key
# Start the server
npm run server
# Or use the built version
npm run build
npm start
# Or install globally
npm install -g agent-mcp-node
agent-mcp --port 8080 --project-dir path-to-directory# 1. Install Agent-MCP
uv venv
uv install
# 2. Start the MCP server
uv run -m agent_mcp.cli --port 8080
# 3. Configure your MCP client to connect to:
# HTTP: http://localhost:8000/mcp
# WebSocket: ws://localhost:8000/mcp/wsFAQs
How do I set up Agent-MCP to coordinate multiple AI agents working on the same codebase without file conflicts?
Clone the repository, configure your API key, then run the Python implementation with your project directory. The framework coordinates agents through shared SQLite-backed memory and task assignment tools enabling dependency checks and progress reporting via view_tasks and update_task_status. Use distinct agent modes like Worker, Research, or Testing to separate responsibilities across parallel instances.
How does Agent-MCP's shared SQLite memory persist context across agent sessions and prevent token limit issues?
Agent-MCP's SQLite backend stores embeddings, task metadata, architectural decisions, and coding patterns as queryable records outside model context windows. When agents restart or hit token limits, they query this external knowledge layer through tools like query_project_context rather than maintaining full conversation history in memory, effectively offloading long-term state to persistent storage that survives process restarts.
What are the risks of using Agent-MCP in a commercial project given its AGPL-3.0 license and maintenance pause since September 2025?
The primary risks are legal compliance and lack of upstream support. AGPL-3.0 requires disclosing modifications if deployed over a network, potentially exposing proprietary logic. The maintenance pause means unpatched vulnerabilities, no compatibility updates for newer MCP specs or dependencies, and no official support. You inherit full maintenance burden for a complex multi-agent system.
How does Agent-MCP compare to Taskmaster MCP for managing tasks across AI coding agents?
Agent-MCP orchestrates tasks across multiple concurrent agents with shared memory and inter-agent communication, while Taskmaster MCP focuses on decomposing and tracking tasks for a single agent without coordination features. Agent-MCP's task tools enable agents to check dependencies before starting work and report progress to other agents, whereas Taskmaster provides sequential task breakdown without multi-agent awareness or shared state synchronization.
What are the best practices for managing multiple MCP servers?
For Model Context Protocol servers, isolate each MCP instance with dedicated ports to prevent transport collisions, use unified authentication tokens post-v1.3.0, and verify transport compatibility since Python SSE and Node.js HTTP/JSON-RPC transports aren't interchangeable. Set environment variables like AGENT_MCP_PORT per instance and monitor endpoint availability through client logs to catch runtime failures early.
What are the main differences between using the Python SDK and the JavaScript SDK for setting up an MCP server?
Python MCP servers typically use stdio transport with frameworks like FastMCP for minimal boilerplate, while JavaScript servers often use Express with HTTP/JSON-RPC requiring manual session management via StreamableHTTPServerTransport. JavaScript leverages TypeScript strict typing for handler safety, whereas Python relies on runtime checks or mypy. Python excels in CLI-style plugins; JavaScript suits web-integrated deployments with persistent sessions.