Verified

AI Brokerage MCP Server

Portfolio review, market context, and order management for brokerage workflows.

Data last scanned today · Reviewed today

01 dangerous tool12 tools0PythonApache-2.0

Overview

This Python-based MCP server streamlines brokerage tasks by simplifying complex API interactions into agent-accessible tools. It focuses on portfolio analysis, position sizing, and order previews, providing a structured approach for AI models to interact with financial data through Interactive Brokers Client Portal-compatible gateways. The server features a demo mode using synthetic account and market data, allowing users to evaluate its workflow capabilities without requiring live credentials. Its architecture incorporates an audit logging system and supports dry-run previews to prioritize visibility and safety during the decision-making process.

Our verdict

The AI Brokerage MCP Server carries a high-risk profile, reflected in its 24/100 security score. This assessment stems from it being a community-maintained project that lacks independent editorial review and has seen limited recent development activity. Users should treat this tool as experimental; it is suitable for testing brokerage workflows in demo mode but requires extreme caution and rigorous manual oversight if configured for live financial transactions.

  • Zero direct dependencies reduce potential supply-chain risks.
  • Built-in demo mode enables safe testing with synthetic data.
  • Includes audit logging features for tracking agent actions.
  • Features real documentation within the project README.
  • Security score of 24/100 indicates high operational risk.
  • Community-maintained status lacks formal vendor verification.
  • Authentication methods have not yet been editorially reviewed.
  • Single contributor increases the bus-factor project risk.

Tools

ToolDescriptionRisk
ibkr_sessionCheck or restore gateway authentication status.low
portfolioAccount totals, positions, concentration warnings, currency exposure, live orders, trades, and recent order history.low
marketTechnical context plus recent broker news headlines for a symbol or option contract.low
search_symbolResolve broker contracts by symbol or company name.low
option_chainInspect option expirations and strikes.low
search_option_contractsResolve exact option contracts.low
get_option_greeksRead option snapshot fields and greeks.low
position_sizeAdvisory risk sizing from account value, stop distance, market data, and exposure.low
submit_orderDry-run preview by default; optionally submit simulated or live orders.medium
manage_orderList, cancel, or modify open orders.medium

Compatibility

ClientLocalDockerRemoteRead-only
Claude Desktop
Cursor
VS Code
Windsurf
ChatGPT

Frequently asked questions

What is the default operational mode of the server?

The server defaults to demo mode, which provides synthetic account, market, option, and order data without requiring any broker credentials.

How do I enable live trading?

To enable live trading, you must set AGENTIC_BROKERAGE_MCP_BROKER_MODE to live and AGENTIC_BROKERAGE_MCP_ENABLE_LIVE_TRADING to true.

What infrastructure is required for live mode?

Live mode requires an active Interactive Brokers Client Portal-compatible gateway, which you must point the server to using the AGENTIC_BROKERAGE_MCP_IBKR_BASE_URL environment variable.

Are order submissions always live?

No, the submit_order tool defaults to a dry-run preview, meaning it will not execute trades unless explicitly configured otherwise.

Does the server store any local data?

Yes, the server uses a local SQLite database to track audit entries, watchlists, and order history.

What system requirements are needed to run the server?

The server is built for Python 3.12 and is managed using the uv package manager as indicated in the setup instructions.

Badge

AI Brokerage MCP Server security score, rated on RepoAI

Maintain this server? Add the live badge to your README.