Overview

tasty‑agent is a Model Context Protocol (MCP) server that bridges large language models with the TastyTrade brokerage platform. It exposes a rich set of tools that let an AI assistant read account data, pull real‑time market information, and even place multi‑leg orders directly from the model’s context. The primary problem it solves is the friction that developers face when trying to give an AI a “hands‑on” trading role: authentication, rate limiting, and complex order logic are all handled behind the scenes. By providing a single, well‑documented API surface, tasty‑agent lets developers focus on building trading strategies and conversational flows rather than wrestling with OAuth tokens or API quotas.

The server’s core value lies in its ability to merge data and action. Account‑level tools such as , , and give the model a complete view of portfolio health, while market‑data tools (, , ) supply the context needed for informed decision‑making. Order management is streamlined through a single call that automatically fetches current prices and supports complex option strategies, all while respecting TastyTrade’s rate limits (2 requests per second). The built‑in dry‑run mode allows safety testing before any live trade is executed.

Key capabilities include:

• Smart order placement with automatic price discovery and multi‑leg support.
• AI‑powered trade idea generation that analyses current positions, watchlists, and market metrics to propose actionable strategies.
• Pre‑built IV rank analysis prompts that surface extreme volatility conditions for entry or exit opportunities.
• Watchlist management to keep the model’s universe dynamic and up‑to‑date.

Typical use cases span automated portfolio monitoring, conversational trading assistants that can suggest or execute trades on demand, and research bots that continuously scan for high‑IV opportunities. In a production workflow, an LLM can query , receive a concise portfolio snapshot, and then invoke to surface new opportunities—all within the same conversational turn. The resulting trade ideas can be passed to for execution or to a human trader for review.

What sets tasty‑agent apart is its tight integration with TastyTrade’s real‑time data streams (DXLink) and the inclusion of advanced option analytics directly in the MCP toolkit. Developers benefit from a single, consistent interface that handles authentication, rate limiting, and complex order logic, enabling rapid prototyping of AI‑driven trading applications without compromising on safety or compliance.