SuperAPI AI Directory

vibetrader-mcp

Trading bot platform: create automated trading strategies in natural language via the Alpaca brokerage.

Open website

About this tool

VibeTrader MCP Server is a Model Context Protocol (MCP) server whose purpose is to connect AI assistants such as Claude Desktop, Cursor, and Windsurf to the VibeTrader trading platform, allowing users to create and manage trading bots, manage portfolios, execute trades, and perform market analysis using natural language. It is officially provided by VibeTrader, and the repository is etbars/vibetrader-mcp.

Main capabilities and available tools

  • Create bots: Create AI-powered trading bots using natural language.
  • Manage bots: Start, pause, delete, and monitor bots.
  • Portfolio: View positions and account balances.
  • Market data: Get quotes and option chains.
  • Backtesting: Test strategies with historical data. The materials state that there are 30 tools in total, and they are available through either endpoint below.

Access methods The hosted server supports two MCP HTTP transports at the same time: /mcp and /sse. Both URLs expose the same 30 tools and use the same authentication method, so existing /sse configurations can continue to be used. The materials recommend that new configurations prefer /mcp, because SSE is deprecated upstream, and /sse is retained only for compatibility with configurations copied from older documentation.

Claude Desktop configuration steps

  1. From https://vibetrader.markets/settings get an API token.
  2. Add it to the Claude Desktop configuration file (~/Library/Application Support/Claude/claude_desktop_config.json).
  3. Clients that connect directly to the remote MCP URL (Cursor, Windsurf, and Claude Desktop's own remote server support) can use only that URL. The materials note that /sse is also available, but new configurations should use /mcp.
  4. Send the token as a header. For the hosted server, this is the authentication method.
  5. Restart Claude Desktop.

Local method It can be run locally via stdio (python server.py stdio), in which case the authenticate tool behaves as expected: provide a token once, and the rest of the requests in this session are authenticated. Local clients can point to http://localhost:8080/mcp or http://localhost:8080/sse。

Authentication notes There is an authenticate tool: under a local stdio configuration, providing a token once authenticates the entire session. However, under HTTP (that is, all the hosted configurations above), the Authorization header should be used instead. The header is parsed per request and is not stored anywhere, so it is more secure: the server does not hold the token at all. Authentication requires an API token; all requests go over HTTPS; the token is parsed per request and is not shared between users—over HTTP, the Authorization: Bearer vt_... header is read only from the user's own request and is not stored at all.

Example conversations

  • Create a momentum bot: "Create a trading bot that buys AAPL when RSI crosses below 30 and sells when it crosses above 70"
  • View bots: "Show me all my trading bots and their performance"
  • Backtest a strategy: "Backtest a moving average crossover strategy on SPY for the last 6 months"
  • Options trading: "Create an iron condor bot for SPY that enters when IV rank is above 50"

Use cases Suitable for users who want to use an AI assistant to create, monitor, and backtest trading strategies in natural language, and to view positions, account balances, quotes, and option chains. Paper trading mode is enabled by default for safety (Paper trading mode is default for safety).

Limitations and information not yet provided The materials do not state specific fees, prices, or subscription plans; they do not state the supported market scope (other than the ticker symbols appearing in the examples), data latency, backtesting accuracy, bot execution frequency, or other details; nor do they state compatibility with clients other than Claude Desktop, Cursor, and Windsurf. The hosted service is deployed on Google Cloud Run at mcp.vibetrader.markets.