Virtual AI Manager (VAM) is an autonomous, multi-agent system that continuously operates to plan, monitor, and execute management tasks. Built around LangGraph orchestration and MCP tool integration, VAM spans a FastAPI backend (agent core) and a Next.js control plane that lets humans supervise and intervene when needed. The MCP tools enable seamless connections to external services (GitHub, Google Calendar, Slack, etc.) so agents can manage calendars, track issues, monitor risks, and report progress without explicit user prompts. The control plane UI provides dashboards, task views, and actionable prompts to review high-risk actions or lifecycle events.

To interact with the MCP server, ensure the backend (FastAPI) is running and the control plane frontend is available. The backend exposes a set of MCP tool servers that agents use to execute actions (e.g., scheduling, authentication, calendar operations, and repository interactions). The system can autonomously decompose goals into DAGs, plan tasks, and monitor execution, while maintaining human-in-the-loop through the RBAC-enabled control plane.

Prerequisites:

• Python 3.10+
• Node.js 18+
• Git
• Internet access to install dependencies
• Optional: GitHub OAuth App and Google OAuth App credentials for calendar/integration

Install steps:

1. Clone the repository git clone https://github.com/agusain2001/Virtual-manager.git cd Virtual-manager

2. Backend setup (FastAPI agent core) python -m venv venv

   Windows

   venv\Scripts\activate

   Linux/macOS

   source venv/bin/activate pip install -r backend/requirements.txt cp backend/.env.example backend/.env # edit with credentials

3. Frontend setup (Next.js control plane) cd frontend npm install

4. Run the system

   Backend (Brain)

   cd backend uvicorn backend.app.main:app --reload

   Backend API runs on http://localhost:8000

   Frontend (Control Plane)

   cd frontend npm run dev

   Frontend dashboard runs on http://localhost:3000

Notes:

• Ensure OAuth credentials (GitHub, Google) are configured in the environment or .env file.
• Update DATABASE_URL to point to a real database if not using the default.
• If you deploy via MCP tooling, you can adjust the mcp_config block to suit your runtime (see JSON in this README).

Tips and common issues:

• Ensure the environment variables for OAuth and Slack are securely stored; do not commit secrets.
• If the backend fails to start, check Python version and dependencies in backend/requirements.txt; run in a virtual environment.
• The MCP tool servers (under backend/app/agents and backend/app/services) expect proper database migrations; run any schema migrations if provided by the project.
• For production, consider configuring a reverse proxy (e.g., Nginx) and a proper DB persisted storage. Adjust HOST/PORT in mcp_config accordingly.
• RBAC is enforced; admins should configure users and roles to control access to sensitive operations.