openprophet.io | learn | github

What is an MCP Trading Server?

An MCP trading server is a program that gives AI models the ability to interact with financial markets. It translates between the language model's reasoning and actual brokerage operations — placing orders, reading market data, analyzing options chains, managing positions.

What is MCP?

MCP (Model Context Protocol) is an open standard created by Anthropic. It lets AI models call external tools. Instead of just generating text, the model can take actions: read files, query databases, call APIs — and in the case of trading — execute buy and sell orders.

The loop:

  1. AI model decides it needs information or wants to act
  2. Sends a structured tool call to the MCP server (e.g., get_quote with symbol AAPL)
  3. MCP server executes the action (fetches quote from Alpaca)
  4. Result returned to model, which uses it in its reasoning

The model chains multiple tool calls to accomplish complex tasks — research a stock, check options chain, analyze technicals, place an order.

How OpenProphet Uses MCP

OpenProphet's MCP server (originally claude_prophet) provides 45+ tools:

Order execution: place_options_order, place_buy_order, place_sell_order, place_managed_position (auto stop-loss/take-profit). All routed through Go backend to Alpaca API.

Market data: get_quote, get_historical_bars, get_options_chain, get_positions. Real-time and historical data for the AI to assess conditions.

News intelligence: get_quick_market_intelligence, get_cleaned_news. Aggregates Google News + MarketWatch, cleans via Gemini. 80-90% noise reduction from raw feeds.

AI memory: find_similar_setups, store_trade_setup. Semantic search over past trades. Local 384-dim embeddings in SQLite via sqlite-vec.

Self-modification: update_agent_prompt, set_heartbeat. The AI modifies its own behavior at runtime.

Other MCP Trading Servers

Why MCP Matters

Before MCP, connecting an AI to a trading system required custom integrations — API wrappers, prompt engineering for structured outputs, fragile parsing. MCP standardizes this into a clean tool interface any compatible model can use.

OpenProphet supports 294+ models from Anthropic, OpenAI, Google, OpenRouter. Same 45+ tools work with all of them because the interface is standardized.

See Also

OpenProphet is experimental software. AI agents can hallucinate and result in total loss of an account. We never recommend connecting real money. This is not investment advice.
CC BY-NC 4.0 · Jake Nesler · Source on GitHub