Claude Fable 5.1 + MCP = New king of Algo-trading!

Summary In this video, Saleh from the YouTube channel Algo-trading with Saleh tests Anthropic’s Claude Fable 5.1 model paired with the Jesse trading framework via MCP (Model Context Protocol). He prompts the autonomous Claude Code agent to research, backtest, optimize, and stress-test an end-to-end algorithmic trading strategy for SPY (S&P 500 ETF) on hourly and 4-hour timeframes, then inspects the resulting backtests, Monte Carlo simulations, generated report, and Python strategy code.

What is shown

• 00:00 - 00:48: Anthropic’s announcement page for Claude Fable 5.1 and Mythos 5.1 (September 2026), showing comparative benchmark tables against Fable 5, Opus 5, and GPT-5.6 Sol, along with a partner quote from Jane Street Capital.
• 01:12 - 02:11: Overview of the Jesse trading framework, Jesse MCP integration with Claude Code, and pricing tiers (Jesse Free Plan, Claude Code Max plan at €90/month, Massive stock data provider).
• 02:12 - 03:02: The prompt entered into Claude Code, requesting an end-to-end research workflow for a SuperTrend long/short strategy on SPY-USD futures with a target Sharpe ratio $\ge 1.5$, 3% account risk, parameter optimization, and Monte Carlo validation.
• 03:23 - 04:06: Review of the initial agent attempt which hit a 1.77 Sharpe ratio but failed to take short trades, prompting a prompt refinement.
• 04:22 - 05:46: Discussion of trading traditional ETF perps on crypto exchanges like Lighter (zero fee DEX) and Hyperliquid (US500-USDC perp), addressing data gaps due to market closing hours.
• 05:47 - 07:27: Final backtest performance metrics on the Jesse dashboard for 2024–2026: 1.90 Sharpe ratio, +59.6% net profit (vs +34.6% buy-and-hold SPY), -9.7% maximum drawdown, 107 total trades, 40.19% win rate, and monthly returns heatmap.
• 07:28 - 08:15: Visualizing trades and indicators on the Jesse interactive candlestick chart (4-hour SuperTrend line, fast EMA, dynamic ATR stop lines).
• 08:16 - 09:19: Validation run across an earlier out-of-sample window (2022–2024) showing +49.4% return, -15.7% max drawdown, and a 1.27 Sharpe ratio.
• 09:20 - 10:22: Monte Carlo candle stress test dashboard across 200 scenarios: original return sits near the median (25.2%), worst 5% at -13.4%, and Sharpe ratio ranging from -0.44 to 2.13.
• 10:23 - 11:06: Full markdown research report auto-generated by the model, detailing objectives, constraints, optimization trials, and recommended next steps.
• 11:07 - 17:19: Code walkthrough in VS Code of SPYUSD_long_short_futures.py, reviewing anchor candle indexing, bull/bear regime filters with ADX, hyperparameter definitions, ATR trailing stops, and execution hooks.
• 17:23 - 19:15: Overview of Jesse’s Community Strategies marketplace and feature roadmap voting dashboard.

Claims & numbers

• The presenter cites Anthropic benchmarks for Claude Fable 5.1: 52.6% on Agentic scientific research (Terminal-Bench Science 0.1[1]), 55.8% (Mythos 5.1: 65.0%) on Agentic coding (Terminal-Bench 4.0), 1853 on Knowledge work (GDPval-AA v2), 77.9% partial / 41.7% strict on Computer use (OSWorld 2.0), 60.9% on Multidisciplinary reasoning (Humanity’s Last Exam), 31.4% on AutomationBench, and 73.4% on Agentic coding (CursorBench 3.2.0).
• The presenter notes Jesse version 3.1.0 added support for stocks, ETFs, currencies, indices, and futures data.
• The presenter mentions using Claude’s Max plan starting at €90 per month.
• The Jesse Discord community is claimed to have more than 5,000 members.
• For the final SPY-USD strategy backtest (2024-09-01 to 2026-08-24):
  • Annualized Sharpe ratio: 1.90 (strategy) vs 1.24 (buy-and-hold SPY).
  • Net profit: +59.6% ($5,958.64) vs +34.6% ($3,460).
  • Maximum drawdown: -9.7% vs -19.3%.
  • Total closed trades: 107 (59 longs / 48 shorts).
  • Win rate: 40.19% (win/loss ratio: 2.66).
  • Maximum underwater period: 133 days.
• In the 2022–2024 prior window check: Sharpe 1.27, net profit +49.4%, max drawdown -15.7% across 126 trades.
• In the Monte Carlo test (200 resampled candle scenarios): original profit was 49.9%, median profit 25.2%, worst 5% loss -13.4%, and best 5% gain 72.7%.

Notable quotes

• 00:35: “In internal benchmarks, Claude Fable 5.1 solves more of our coding problems than Fable 5 or Opus 5, and achieves state of the art on trading intuition.” (Craig Falls, Head of Quantitative Research at Jane Street Capital, quoted by the presenter).
• 03:23: “Look at that. It almost one-shot the whole thing.”
• 10:40: “Like, this is really complete. Like, if this was an actual person that you gave it the task to go and do research for you, you can imagine this was the results that they gave you back...”

Assessment This is a real community demonstration and review of Claude Fable 5.1 using Claude Code via MCP to automate quant research in the Jesse trading framework. The backtest results, interactive charts, terminal logs, and generated strategy code are shown directly in real application interfaces, though the lengthy iteration and optimization phases were completed off-camera and shown as finished runs.

Described by gemini-3.8-flash on 2026-09-29 from the video’s audio and frames.