📈 TradingCLI

A fast, local-first paper-trading and market-simulation toolkit for your terminal.

TradingCLI gives you a fake brokerage account that lives in one file on your own machine. You place orders against it, and it fills them using real market prices from Yahoo Finance. No signup, no API key, and no code path that reaches a real broker.

        you  ·  or your AI agent
                    │
                    ▼
     tradingcli  (dashboard · commands · MCP server)
                    │
       ┌────────────┴────────────┐
       ▼                         ▼
 ~/.papertrade.db          Yahoo Finance
 accounts, orders,         prices only,
 positions, ledger         read-only

🚀 Install and run

Requires Python 3.10 or newer.

python3 -m venv .venv
source .venv/bin/activate
python3 -m pip install .
tradingcli

That last command opens the dashboard and a first-run setup wizard. From there:

tradingcli buy AAPL 5        # place a trade
tradingcli positions         # see what you hold
tradingcli backtest          # test today's holdings against history
tradingcli --help-all        # every command, no market data needed

Three ways to drive it — all sharing the same database:

Related MCP server: tastytrade-mcp

🖥️ Screenshots

The captures come from a local paper-trading database; they contain no real portfolio or credential data.

✨ What you get

🤖 For AI agents

The usual way to let an agent practise trading is a hosted paper API such as Alpaca's: signup, API keys, and a network round trip per order. TradingCLI does the same job with a local SQLite file — no keys to leak, no rate limits, and prices you can pin so a run replays exactly, which a live API cannot do.

PAPERTRADE_DB=./agent-sandbox.db tradingcli-mcp

→ docs/agents.md — the full comparison, the safety rails (narrow tool catalog, idempotency keys, engine-enforced risk limits, audit trail), and how to wire up an MCP client.

📚 Documentation

📊 Backtesting

Press g in the dashboard to open Backtesting & Graphs. It puts the selected portfolio's live performance curve next to a backtesting.py backtest of the same holdings, reporting return, CAGR, volatility, Sharpe, Sortino, costs and maximum drawdown. The universe comes only from that account's open positions; another account's names never leak in. Pick a window on open (6m, 1y, 2y, 5y, 10y, max, or an exact number of days).

🛡️ Safety

TradingCLI never places live brokerage orders, and Yahoo Finance is not an exchange-grade feed. Do not use it as the sole source for financial decisions, and do not expose its stdio MCP server as an unauthenticated network service.

Fetching prices, news, history and option chains sends held, pending or watchlisted symbols to Yahoo Finance through yfinance. See SECURITY.md for the local-data and network privacy model.

🧑‍💻 Development

uv sync --locked --extra dev
uv run python run_tests.py
ruff check .

The runner discovers every test_*.py contract and executes each one in an isolated subprocess and temporary database. GitHub Actions runs the same checks on Python 3.10 and 3.13.

Contributions are welcome — see CONTRIBUTING.md. Security reports should follow SECURITY.md, not a public issue.

📄 License

Released under the MIT License.