agent-hub

Agent Hub is an MCP-based orchestration platform that coordinates a multi-agent ecosystem to automate developer workflows, security checks, data processing, and cloud operations. Powered by FastMCP with a Python runtime and JWT-based authentication, it exposes 50+ tools through a unified MCP interface and delegates tasks to specialized AI agents (GitHub, Azure, Security, PDF, Scraper, Chart, and more). Users can query the system to perform repository operations, security scans, data extraction, and complex multi-step workflows, all orchestrated by the MCP server and driven by the API gateway. To use it, start the MCP server, connect to the API gateway, and issue queries that trigger dynamic tool composition, agent selection, and tool execution across the agent ecosystem. The platform is designed to run in cloud-native environments and can integrate with GitHub, Azure services, Snyk, and external APIs for end-to-end automation.

Once running, you can leverage the Nine specialized AI agents to perform targeted tasks, such as GitHub operations (Repository Management, Issue/PR automation), Azure resource orchestration, Security vulnerability scanning and remediation guidance, PDF and document processing, web data extraction, and data visualization. The MCP tooling layer (FastMCP) acts as the extensible backbone, enabling 50+ tools and seamless integration with external services. Use the API gateway to route user queries to the appropriate agents and tools, allowing for real-time orchestration and context-aware responses.

Prerequisites:

• Python 3.12+ (and virtual environment support)
• Network access to install dependencies
• Optional: Docker for containerized deployments

1. Prepare a Python virtual environment

   • python3.12 -m venv venv
   • source venv/bin/activate (Linux/macOS)
   • venv\Scripts\activate (Windows)

2. Install required dependencies (via pip)

   • pip install -r requirements.txt
   • If a requirements file is not provided, install core MCP dependencies a la carte, e.g.:
     • pip install fastmcp fastapi uvicorn pydantic jwt

3. Configure environment variables

   • Create a .env file with required credentials and settings (example placeholders):
     • MCP_JWT_SECRET=your_jwt_secret
     • MCP_API_BASE_URL=https://your-api-gateway.example.com
     • AZURE_OPENAI_API_KEY=your_azure_openai_key

4. Run the MCP server

   • python -m fastmcp_server
   • Or specify a module entry if using a different entry point, e.g.:
     • python -m fastmcp_server --config /path/to/config.json

5. (Optional) Run in Docker

   • Build: docker build -t agent-hub-fastmcp .
   • Run: docker run -e MCP_JWT_SECRET=your_jwt_secret -p 8000:8000 agent-hub-fastmcp

6. Verify installation

   • Access the MCP server API at http://localhost:8000 (or configured base URL)
   • Ensure authentication works and tools are discoverable via MCP client tooling.

Notes and tips:

• The MCP server relies on a JWT-based authentication flow; securely manage your secret keys.
• When integrating with Azure OpenAI and other services, ensure API keys and access tokens are kept secure and rotated regularly.
• If you encounter tool discovery issues, verify that the MCP tool registry includes the 50+ tools and that agents (GitHub, Azure, Security, PDF, Scraper, Chart, etc.) are properly registered and reachable.
• For production deployments, consider running multiple instances behind a load balancer, and enable TLS termination at the API gateway.
• Logging and observability are important; enable structured logs and, where possible, integrate with your existing SRE tooling for metrics and tracing.