The Academy is an MCP-enabled agent orchestration platform that lets you run multi-agent dialogues, conduct bulk experiments, and analyze conversation data through a comprehensive set of MCP tools. The MCP API exposed at /api/mcp provides 70+ tools organized into resources, sessions, participants, conversations, analysis, exports, and AI provider integrations. You can manage sessions, add participants (AI models), start and pause conversations, send messages, run analyses, and export results all through the MCP interface or via JSON-RPC calls. Start by creating a session, then add participants such as Claude, GPT, Gemini, Grok, and others, configure analysis settings, and launch automated conversations or bulk experiments. Use the analyze and export tools to surface insights and generate datasets for study.

To execute operations, call tools like create_session, add_participant, start_conversation, analyze_conversation, and export_session. Bulk experiments can be created and run with create_experiment and execute_experiment, then you can fetch results with get_experiment_results. The platform supports live analysis, multiple providers, and a range of control tools to pause, resume, or inject moderator prompts during conversations. You can inspect individual sessions, their messages, and associated analysis snapshots via the MCP resources (academy://sessions, academy://session/{id}, academy://session/{id}/analysis, etc.).

Prerequisites

• Node.js 18+ (for the server)
• Docker (optional, for PostgreSQL or local deployment of dependencies)
• Access to API keys for the providers you plan to use

Install and run locally

1. Clone the repository git clone https://github.com/yourname/the-academy.git cd the-academy

2. Install dependencies npm install

   or if you use pnpm

   pnpm install

3. Prepare environment variables Create a local environment file or set env vars in your shell or deployment Example (dotenv/.env.local): ANTHROPIC_API_KEY=your_key OPENAI_API_KEY=your_key XAI_API_KEY=your_key GOOGLE_AI_API_KEY=your_key DEEPSEEK_API_KEY=your_key MISTRAL_API_KEY=your_key COHERE_API_KEY=your_key DATABASE_URL=postgresql://academy_user:academy_password@localhost:5432/academy_db PORT=3000

4. Start the database (if using Docker) docker-compose up -d

5. Run the server npm run dev

   or if using pnpm

   pnpm dev

6. Verify Visit http://localhost:3000 and/or use the MCP API at http://localhost:3000/api/mcp

Notes

• If you already have a PostgreSQL instance, update DATABASE_URL accordingly.
• Ensure all desired provider keys are configured before starting experiments.

Tips and common issues:

• Ensure your API keys are valid for the providers you plan to use; missing keys will cause tool calls to fail.
• The MCP tools can be invoked via a JSON-RPC interface; use mcp.callTool('<tool_name>', { ...params }) to interact programmatically.
• For bulk experiments, define clear templates and experiment configurations to avoid runaway sessions.
• Check environment-specific differences (local vs. production) for database connections and port mappings.
• If you encounter API errors, use get_api_errors and log_api_error utilities to diagnose and retry operations.
• The platform exposes 70+ tools; consult the resources and tooling sections in the README for exact tool names and expected payloads.