- Home
- Skills
- Trading & Investing
- QuantDesk — AI Quant Research & Backtesting Agent
QuantDesk — AI Quant Research & Backtesting Agent
Run reproducible crypto backtests, compare strategy configurations, and test paper-order risk controls locally from your AI coding agent. Export Markdown, JSON and CSV with the exact inputs, costs and assumptions.
$5
Works with the AI Agents you already use
QuantDesk — AI Quant Research & Backtesting Agent
Example session with this skill installed
Using my Binance BTC/USDT one-hour candles for January 2024, backtest an EMA 9/21 crossover with 10,000 USDT starting cash. Charge 10 basis points of fees per fill and assume 5 basis points of adverse slippage. Export the run and report the return, risk and costs.
- Read your context and instructions
- Compiled the quantdesk
- Generated the data export
QuantDesk backtest
BTC/USDT spot · 1h · 744 bars · Binance Public Data
From 2024-01-01T00:00:00Z through 2024-01-31T23:59:59.999Z.
CSV SHA-256 01fe03aa2bb3949031552f93d992dfe294e7a5dabe673ad2eab5a87d7d548edf.
| Metric | Observed result |
|---|---|
| Ending marked equity | 9,520.41 USDT |
| Total return | −4.80% |
| Annualized return | −43.94% |
| Sharpe | −5.85 |
| Sortino | −7.35 |
| Maximum drawdown | 5.59% |
| Closed-trade win rate | 17.65% |
| Closed round trips | 17 |
| Fees | 154.03 USDT |
| Modeled slippage | 77.01 USDT |
| Total modeled transaction costs | 231.04 USDT |
| Turnover | 15.78× |
Run settings
EMA periods 9 and 21, ATR period 14, minimum trend strength 0. Starting cash 10,000 USDT. Fee 10 bps and adverse slippage 5 bps per fill. Risk per trade 0.5%, daily realized-loss threshold 2%, max loss streak 3, max one open position. ATR stop and take-profit multipliers 1.5 and 3. Simulated step size and minimum quantity 0.000001 BTC, minimum notional 1 USDT, and tick field 0.01 USDT. Fixed-bar annualization uses 8,760 hourly periods.
Signals consume closed bars and orders use the next open. Intrabar protection uses OHLC assumptions. The annualized result extrapolates a 31-day historical simulation. Short samples and sparse trades limit inference.
Files and reproduction
Generated report.md, result.json, metrics.csv, trades.csv and equity.csv. Re-run backtest with the same input hash and settings in a new output folder. This recorded run used runtime 0.1.0 and Python 3.11.5 on macOS ARM64.
metrics.csv
CSV · data export
Example file from a real run - the skill writes it into your workspace.
Connects securely to your tools. The creator never sees your data.
About this skill
Turn a request like “Backtest a 9/21 EMA crossover on this BTC/USDT CSV and report drawdown and trading costs” into an executable QuantDesk run. Your agent validates the supplied market bars, runs the local engine and reports what the simulation found.
What you can run
- Backtest the built-in EMA crossover and RSI/Bollinger mean-reversion strategies on fixed-interval crypto spot candles.
- Compare two to six configurations on the same bars, starting cash, position risk and fees. Inspect total return, Sharpe, Sortino, drawdown, win rate, turnover and costs.
- Test virtual orders against cash, inventory and position limits. Inspect simulated fills and rejected intents in a local SQLite ledger.
- Export Markdown, JSON and CSV, including the input file digest, strategy settings, modeled fees and slippage, and runtime identity.
Setup
Use a terminal-capable Agensi-connected agent and Python 3.11+ with SQLite. Download and verify the QuantDesk runtime wheel, version 0.1.0. Install the downloaded wheel in the Python environment the agent uses. The runtime processes your named inputs offline. Runtime wheel and SHA-256
python -m pip install --no-index --no-deps quantdesk_agensi_runtime-0.1.0-py3-none-any.whl
Scope
This release models single-symbol, long-only crypto spot trading on up to 2,000 consecutive bars per run. The two registered strategies use next-open execution and explicit OHLC assumptions. Short samples can yield unstable metrics. Paper orders use virtual balances. Supply a local OHLCV CSV or run the included synthetic sample.
Find more skills like this on the Claude Skills Marketplace, the Codex Skills Marketplace and the Gemini Skills Marketplace, or browse Claude Skills for Trading & Investing.
Single-symbol, long-only crypto spot research. Two built-in strategies, seven fixed intervals and up to 2,000 consecutive bars per run. OHLC-based fills model execution assumptions, and gaps can exceed nominal per-trade risk. Annualized results over short windows can be unstable. Paper scenarios use virtual orders and a fresh local ledger. Live trading, automatic market data fetching and equity calendars are outside this release.
How to use
- 1
Connect your agent to the Agensi MCP
Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup
- 2
Sign in to Agensi
Your agent shows a sign-in link. Sign in and click Allow.
- 3
Ask for what you need
Your agent loads this skill and uses it on your task. It always asks before buying a skill.
Reviews
No reviews yet
Be one of the first to try it. Every listed skill goes through the trust checks below.
Security scanned
Checked by our 8-point scan before listing
Fresh listing
Recently published to Agensi
Money-back guarantee
Doesn't do what its page promises? You get your money back.
Trust & safety
Security scan
Security scan passed 1 day ago
- Software
- Python 3.11+ with SQLiteQuantDesk Agensi runtime wheel 0.1.0Terminal-capable Agensi-connected agent
Python
Install the verified QuantDesk runtime wheel from the Documentation link in a dedicated Python environment. Run the Skill using that same interpreter. The included synthetic demo runs offline.
Permissions
The terminal runs the packaged Python adapter. The runtime reads only the named local JSON and OHLCV CSV inputs and writes reports or a fresh paper SQLite ledger to a new output directory. No exchange connection or credential file is required.