composer-trade

Composer-trade MCP exposes a Python-based MCP server that lets LLMs like Cursor or Claude validate investment ideas, backtest strategies, and run multiple investment symphonies in parallel to compare live performance. The server connects to Composer services (backtesting, strategy database, and live monitoring) and provides a structured MCP interface so your agent can request backtests, retrieve performance metrics, and prompt the system to rebalance or adjust strategies. Use the included prompts and examples to ask for strategy building, backtest comparisons, or best-performing symphonies, then let the system run in parallel across multiple symphonies to identify top performers.

To start using, install the MCP server, ensure your environment is configured with your Composer API key, and run the server. Connect your MCP-enabled LLM client to the server URL exposed by this MCP instance. You can then issue requests like:

• Build and backtest a strategy using RSI and EMA indicators across supported assets.
• Compare two symphonies and plot their performance metrics side by side.
• Identify your best-performing symphonies and analyze why they are working (risk-adjusted return, drawdown, Sharpe ratio, etc.).

Prerequisites:

• Python 3.8+ installed on your system
• internet access to install dependencies
• Composer account for API access (subscription may be required for live trading features)

Install steps:

1. Create and (optional) activate a virtual environment: python -m venv venv source venv/bin/activate # macOS/Linux venv\Scripts\activate # Windows

2. Install the MCP package from PyPI: pip install composer-trade-mcp

3. Verify installation (example, adjust if your entry point differs): python -m composer_trade_mcp --version

4. Start the MCP server (use your API key in the environment if required): export COMPOSER_API_KEY=your-composer-api-key # macOS/Linux set COMPOSER_API_KEY=your-composer-api-key # Windows python -m composer_trade_mcp

5. Connect your MCP client (LLM) to the server URL that is exposed when the server starts.

Notes:

• If you prefer not to rely on a global Python installation, consider using a virtual environment or a tool like pipx to isolate the MCP server.
• Ensure network egress to Composer services is allowed in your environment.

Tips and considerations:

• You will typically need a Composer API key to enable live trading and backtesting capabilities.
• If you plan to run multiple symphonies in parallel, consider increasing concurrency limits in your environment and ensuring your network can handle the load.
• The MCP server may require specific environment variables for API endpoints or authentication; consult the Composer docs for current variable names and values.
• For production usage, consider running the MCP server behind a lightweight reverse proxy and enabling authentication at the MCP client level.
• If you encounter rate limits or API errors, check your API key validity and ensure you’re within your subscription plan’s allowances.
• The example config uses a single server named “composer”; you can add more servers or customize the server name as needed.