Carmenta is a Node.js-based MCP server that combines memory-enabled conversations with a rich, interface-driven AI experience. It brings memory across sessions, voice-first interactions, model selection, and an extensible agent framework into a single subscription. The server exposes an interface that coordinates memory management, AI agents, and service integrations to provide proactive, contextual assistance rather than simple chat replies. You can leverage built-in capabilities like memory persistence, voice interactions, agent scheduling, and browser automation through a unified API and UI layer. You’ll likely interact with the server through the web UI and its AG-UI protocol-driven components which return structured results (maps, reports with citations, task actions) instead of plain chat bubbles. The system is designed to let you pick models or adjust a speed/quality slider while the preprocessing layer handles model routing, prompt enhancement, and response strategy automatically.

Prerequisites:

• Node.js 24+ installed on your system
• pnpm 10.x+ or npm (depending on your preferred workflow)
• Git to clone the repository

Step-by-step:

1. Clone the repository git clone https://github.com/nicholaswilliams/carmenta.git cd carmenta

2. Install dependencies pnpm install

   or if you prefer npm:

   npm install

3. Configure environment (optional but recommended)

   • Create a .env file at the project root with your preferred settings (e.g., memory backend, API keys, service integrations).
   • Example placeholders: MEMORY_STORE=redis://localhost:6379 OPENAI_API_KEY=your-key-here CLARA_API_KEY=your-key-here

4. Start the development server pnpm dev

   or with npm:

   npm run dev

5. Run tests, type checks, and lint as needed pnpm test pnpm type-check pnpm lint

6. Access the server Open http://localhost:3000 in your browser (default port may vary based on config).

Notes:

• The project uses Next.js 16 with TypeScript and a modular knowledge-driven architecture. The knowledge/ directory contains the product specifications that drive code generation.

Tips and considerations:

• The Memory feature is central to Carmenta; ensure your memory store backend is configured and reliable for persistent conversations across sessions.
• If you plan to deploy, consider using a production-grade reverse proxy and a persistent database for logs, memory state, and agent data.
• The system includes an AI Team with proactive agents; you can customize roles (Researcher, Analyst, Creator, Reviewer) to fit your workflow.
• For model selection, you can leverage the built-in 1x/10x/100x framework to balance speed and quality according to the task.
• When integrating external services (storage, calendars, authentication), ensure proper scoping of API keys and least-privilege access.
• If encounters issues with memory loss or misrouted prompts, check the knowledge/architecture and memory configuration to confirm proper context retention.