LLMs Get a Brokerage Account

What it does This is an official Model Context Protocol server for Alpaca’s trading API. It translates natural language requests from LLM clients like Claude or Cursor into actual market operations—buying stocks, ETFs, crypto, and options, pulling real-time quotes, and checking portfolio history. Under the hood, it is essentially a structured bridge between conversational AI and a brokerage backend.

The interesting bit The project treats an LLM as a literal trading interface, which is either convenient or terrifying depending on your risk tolerance. To mitigate the latter, it defaults to paper trading and supports server-side toolset filtering so you can, for example, let the model read market data while explicitly denying it permission to exercise options.

Key highlights

• Covers the full Alpaca surface area: stocks, ETFs, crypto, options (including multi-leg strategies), watchlists, corporate actions, and news
• Server-side toolset filtering via ALPACA_TOOLSETS lets you restrict which API capabilities the LLM client discovers
• Defaults to paper trading; live trading requires explicitly setting ALPACA_PAPER_TRADE to false
• V2 is a complete rewrite built on FastMCP and OpenAPI, generating tools directly from Alpaca’s API specs rather than hand-crafting schemas

Caveats

• Zero backward compatibility with V1: every tool name, parameter, and schema changed, so existing workflows and prompts will break
• Configuration is strictly environment-variable based inside your MCP client; there is no standalone config file or CLI setup step

Verdict Worth a look if you want to experiment with algorithmic strategies through conversational prompts, or if you live in Cursor and hate switching tabs to check positions. If you do not use an MCP-compatible AI assistant, this is just a very indirect way to call a REST API.