Open-source finance research agent · 🇰🇷 Korean market fork

Turn Korean market questions into runnable research.

Vibe-Trading-KR is the Korean-market fork of Vibe-Trading — natural-language research over market data, backtests, and swarm analysis, plus Korean broker connectors (KIS · LS · DB · Kiwoom) and KRX backtesting.

vibe-trading

$ vibe-trading run -p "Backtest 005930.KS 20/50 MA for 2024"

route: kr_equity data -> strategy -> backtest -> run card

return +18.6% max drawdown -7.4%

$ vibe-trading --swarm-run investment_committee

workers: macro, quant, risk, catalyst

report saved with evidence, metrics, and caveats

Install

One command to start a research workspace.

# upstream core only (no KR connectors)
pip install vibe-trading-ai
vibe-trading init
vibe-trading

🇰🇷 Korean market

Korean broker connectors & KRX backtesting.

Broker connectors

KIS · LS · DB · Kiwoom REST registered as broker_sdk profiles; Kiwoom OpenAPI+ and Daishin CYBOS as Windows local bridges. Status is tracked conservatively — real-account orders stay behind mandate, kill switch, and audit gates.

Broker API matrix →

KRX backtesting

005930.KS, 035720.KQ, KRX:005930 route to kr_equity. KRX 1-share lot and ±30% daily limit modeled; benchmark resolves to KOSPI ^KS11.

Credentials & safety

Broker keys come from environment variables; per-connector config is stored under ~/.vibe-trading with 0600 permissions. Live order paths require credentialed smoke evidence first.

Credential smoke runbook →

Wiki

Explore Vibe-Trading-KR.

Research loop

Every answer keeps the trail inspectable.

01

Route

Select skills, data sources, tools, and swarm teams from the request.

02

Ground

Fetch market data, documents, URLs, broker journals, or local files at runtime.

03

Test

Run backtests, factor analysis, options checks, validation, and report generation.

04

Deliver

Return run cards, exports, reports, tool traces, and caveats for the next session.

Capabilities

Built for finance research, not brokerage execution.

Self-improving trading agent

Agent Harness

Persistent memory, session search, editable skills, context compression, and tool-batched research runs.

Cross-market data and backtesting

Backtesting

Seven engines across equities, crypto, futures, forex, composites, and options portfolios.

Multi-agent trading teams

Swarm Teams

Investment committee, quant desk, risk review, macro, crypto, and other preset research teams.

Shadow Account workflow

Shadow Account

Parse broker journals, extract implicit rules, run counterfactual backtests, and render audit reports.

Boundary

No live trading.

Vibe-Trading is for research, simulation, and backtesting. It does not execute live trades and does not provide investment advice.