What matters most when vibe trading FX?

Letting the numbers, not the AI, decide. kfxai put its direction-prediction model through walk-forward testing over roughly 10,000 trades and discarded it once the data showed no edge. Throwing away a confident-sounding proposal that fails validation is the core discipline of vibe trading.

Is real money at risk?

No. kfxai paper-trades against live prices from an OANDA demo environment; no real funds move. This site is not investment advice and implies no endorsement by OANDA.

PAPER TRADING · OANDA PRACTICE · EVERY NUMBER PUBLISHED

An FX bot that killed its own prediction model — and trades the strategy that survived.

Q: What is vibe trading?

Vibe trading means describing a trading strategy to an AI in plain language, letting the AI write the code, and deciding whether to keep it based on backtest numbers. Humans supply the hypothesis, the AI supplies the code, and the data decides — the trading counterpart of vibe coding.

kfxai is the FX sibling of kfreqai, run by the same jellyfish AI VTuber, Kurage. Before letting it trade, we built a walk-forward backtest harness and measured ML direction prediction across ~10,000 simulated trades. The verdict: no edge, at any timeframe. So we threw it away — and adopted the one strategy that survived every robustness test: a Tokyo-range → London-breakout session strategy on JPY pairs.

Read the Trading Diary Dashboard GitHub Crypto sibling: kfreqai

Kurage 🪼

AI VTuber · runs the bot on OANDA Practice, and publishes what the data actually said.

What is vibe trading? — Growing an FX strategy by talking to an AI

Writing code by talking to an AI is called vibe coding. In the same spirit, describing a trading idea in plain language, letting the AI turn it into code, and growing the strategy by judging it on backtest numbers is what we call vibe trading. kfxai is that process applied to FX — including the parts that failed.

Ask for a strategy in plain language

“Write a breakout strategy that goes long when the Tokyo-session range breaks, stop at half the range width.” That one sentence to Codex or Claude Code produces working strategy code. No programming required to start.

The data decides, not the AI

The discipline that makes vibe trading work is refusing to trust the model. We ran direction prediction through walk-forward testing over ~10,000 trades and discarded it once the data showed no edge.

Only survivors go live

Session breakout stayed profitable across every parameter set over 1.6 years, so that is the only thing we trade. Nothing reaches production without surviving validation.

The rejections are public too

The backtest harness ships in the repo, so you can see what was tried and what failed. The record of failures is what fuels the next hypothesis.

Vibe trading toolkit guide Try it on crypto (kfreqai) GitHub

Built in the honest order

Measure first, trade second. The backtest harness ships in the repo.

DISCIPLINE

We killed direction prediction with data

Logistic regression and LightGBM, on M15 / H1 / H4 candles, with triple-barrier labels and richer features — 7 configurations, ~10,000 walk-forward trades. Direction accuracy came out at 49–52% against a 52.6% breakeven. Every configuration lost money. The harness that proved it is scripts/backtest.py, in the open.

LIVE

Session breakout — the survivor

Tokyo range (00–07 UTC) broken during London hours (07–12 UTC), traded with the trend: stop at the far side of the range, target at 1.5× range height, forced flat at 21:00 UTC, one trade per pair per day. Validated over 1.6 years / 570 trades — positive on every take-profit multiple and buffer we swept.

LIVE

JPY pairs only — for a reason

The edge concentrates in USD/JPY, EUR/JPY and GBP/JPY — where the Tokyo session actually sets a meaningful range. AUD/JPY lost in both test windows and EUR/USD flipped sign between parameters, so both were excluded. Structure first, not cherry-picking.

CORE

OSS body with fixed risk gates

OANDA v20 connectivity, spread cap, daily-loss cap, max positions, stop/target on every order, SQLite trade journal, and a live dashboard. GPL-3.0 — clone it and it runs in paper mode with a rule-based brain, no LLM required.

CORE

Trust boundary: the brain can't touch money

The judgment layer (market regime, risk directives, trade reviews) is pluggable — rule-based, local Gemma 4 via Ollama, or a paid x402 API. Whichever you pick, it holds no broker credentials, decides no order sizes, and cannot place orders. Final risk decisions live in the body you run yourself.

NEW

Metered brain: Kurage FX Brain

The paid judgment API vendors the LLM intelligence of popular open-source trading projects — TradingAgents (Apache-2.0), FinGPT (MIT), ai-hedge-fund (MIT), pinned commits, full attribution. Paid calls use DeepSeek; self-hosted deployments can use local Gemma. The catalog contains 34 skills across 36 live x402 paths at $0.05 per standard call, plus the full multi-agent debate graph at $0.50. USDC on Base, JPYC on Polygon, or Bankr.

What the data said

The numbers below are why the bot trades what it trades.

H4 candles, 2.5 years, 2,922 walk-forward trades: direction accuracy 49.2% — a coin toss. H1 and M15 did no better, with either model. A strategy that only fires when the model is confident still lost on every window. So kfxai does not trade ML direction predictions.

One "edge" looked real: AUD/JPY made +7,277 (paper JPY) on H1 — then flipped to -26,697 on H4. Sign instability across windows and timeframes is how model variance disguises itself as signal. Excluded.

The session breakout survived everything we threw at it: two separate windows, a take-profit sweep that rose monotonically from +770 to +8,841 (paper JPY, 1,000 units), buffer sweeps all positive, and a structural story — the Tokyo range means something for JPY pairs. That's the one running on paper now.

Live positions & cycles Read the code

OSS body, Kurage FX Brain

The broker-facing body is open source. The LLM intelligence is optional and metered.

OSS BODY

kfxai owns execution and risk

OANDA connectivity, order sizing, stops, daily-loss limits and the trade journal stay in the GPL-3.0 body. Kurage FX Brain only returns analysis and judgments; it has no broker credentials and no order-execution endpoint.

34 SKILLS

The complete FX intelligence catalog

Technical, macro and sentiment analysis, trade/risk/portfolio decisions, FinGPT, FinRobot, FinMem, market rankings and a complete TradingAgents multi-agent graph are documented with request prices on the dedicated product page.

Open Kurage FX Brain →

ACCESS

Direct x402, Bankr and JPYC

Direct Coinbase CDP x402 starts at $0.05 per call, Bankr exposes the core skills, and JPYC provides a Polygon payment rail. RapidAPI products are configured in Studio, but public marketplace listings are still pending.

View live x402 discovery →