QuantDinger (brokermr810)

by brokermr810

Community Data Science & ML 10k likes

The founder-maintained mirror of QuantDinger, a self-hosted AI trading OS whose MCP server gives Cursor, Claude Code, or Codex scoped access to markets, backtests, and paper/live trading.

Install the MCP server

pip install "quantdinger-mcp==0.5.0"

Source: https://github.com/brokermr810/QuantDinger

About

This is the same open-source QuantDinger codebase published under OpenByteInc, hosted here on the project founder's personal account: a Python-first, self-hosted trading OS covering strategy authoring, backtesting, paper trading, and live execution across crypto exchanges (Binance, OKX, Bybit, Gate, HTX, Bitget) and brokers like IBKR and Alpaca, with Prometheus/Grafana monitoring and Postgres-backed audit logs.

The included quantdinger-mcp server wraps QuantDinger's Agent Gateway, not the exchanges themselves, so an AI client gets tools for reading markets, managing watchlists and strategy versions, running backtests, and observing broker state read-only, while credentials and admin access stay behind the gateway. Live order placement is opt-in at the server level and requires an authorized agent token plus an idempotency key on every mutating call.

Key features

  • Python Strategy API V2 with versioned indicators and strategies
  • Server-side backtesting and paper trading before committing real capital
  • Live execution across multiple crypto exchanges plus IBKR/Alpaca
  • MCP server that proxies the Agent Gateway instead of exposing exchange keys directly
  • Postgres audit logging with Prometheus/Grafana metrics and alerting
  • One-command Docker-based installer for self-hosting

Use cases

  • Having an agent evaluate backtest results and suggest parameter changes before a strategy goes live
  • Running unattended paper trading while an agent watches for signal alerts
  • Reviewing a strategy's deployment and version history through agent tool calls
  • Keeping full custody of exchange credentials while still letting an AI client drive research and backtests

Available tools

market_and_watchlist_tools

Symbol discovery, OHLCV history, and watchlist/universe management for research.

indicator_tools

Computes indicator series for charting and strategy inputs.

strategy_tools

Compiles, saves, and version-tracks Strategy API V2 code.

backtest_and_job_tools

Starts backtests and polls or streams their status and results.

deployment_tools

Creates strategy deployments in a stopped state pending manual activation.

broker_and_alert_tools

Read-only broker/execution observation plus scoped signal-alert notifications.

Frequently asked questions

Is this a different project from OpenByteInc/QuantDinger?

No. It's the same codebase; brokermr810 is the founder's personal GitHub account hosting an identical copy of the QuantDinger repository.

Can the MCP server execute live trades by itself?

Only if the server operator has explicitly enabled live trading and authorized the agent token. By default execution runs in paper mode and every mutating call needs an idempotency key.