0bps

From spec to running system

This site gives you 5,806 strategy specs and the theory behind them. What it can’t give you is the other 90% of a trading system: execution, state, risk gates, tests. That is what AgenKit is for.

What a spec gives you

  • Entry and exit rules
  • Indicator maths
  • Default parameters
  • A backtest someone ran once

What a live system needs

  • Order lifecycle with retries and idempotent order IDs
  • Position state that survives a crash and reconciles with the exchange
  • WebSocket feeds that reconnect and backfill gaps
  • Hard risk gates: size caps, daily-loss halt, kill switch
  • Tests proving the indicator matches the spec, bar for bar
  • Paper trading by default, live only by explicit opt-in
  • Secrets handled safely, rate limits respected

How /agenkit runs it

You describe the task once, inside the coding tool you already use. A team of specialist agents takes it phase by phase and pauses for your approval at each step.

  1. 1

    Orient Context engineer

    Indexes your project and the spec so later steps work from what is actually there.

  2. 2

    Brainstorm Conductor

    Turns “build this strategy” into a clear spec: venues, instruments, risk limits, what “done” means.

  3. 3

    Architecture Backend architect

    Order lifecycle, state, idempotency, data feeds, reconnection, secrets.

  4. 4

    Build, test-first Test engineer

    Indicator maths tested bar-for-bar against the spec, then the implementation.

  5. 5

    Review Code reviewer

    Checks the build against the spec first, then for code quality and security.

  6. 6

    Ship Infra / DevOps

    CI, deploy, verification — with paper trading on until you switch it off.

Start in two minutes

# in your project folder (Node.js 18+)
npx agenkit activate <license-key>
npx agenkit install engineering-kit

# optional: index a clone of the strategy vault so specs are found cheaply
git clone https://github.com/brainbrick-trades/The-Quant-Trading-Vault
cd The-Quant-Trading-Vault && npx agenkit memory build

Then, in Claude Code, Cursor, Codex or any of the 7 supported tools, type /agenkit followed by what you want. Every strategy page on quants.md has a command pre-written for that strategy.

Ready-made commands

Five starting points from the vault, each a different kind of system.

Multi-asset momentum

/agenkit build a production research system from strategies/Python-Version-Multi-Asset-Momentum-Strategy-Tutorial.md

Multi-symbol ATR futures bot

/agenkit implement strategies/Digital-Currency-Futures-Multi-Variety-ATR-Strategy-Teaching.md as a research backtest with delay-1 execution and fee modeling

Adaptive grid bot

/agenkit implement strategies/Adaptive-Intelligent-Grid-Trading-Strategy.md as a production grid bot for Hyperliquid perps with order deduplication and inventory controls

Intertemporal arbitrage

/agenkit build a spread arbitrage engine from strategies/High-Frequency-Intertemporal-Arbitrage-Strategy.md with WebSocket order routing

Avellaneda–Stoikov market maker

/agenkit build an Avellaneda-Stoikov market making bot for perps from strategies/Dynamic-Spread-Market-Making-Strategy.md with inventory skew management

Ship it like a team would.

One subscription, every supported coding tool, updates included. See current pricing on the AgenKit site.

Get AgenKitRead the docs

Disclosure: AgenKit is built by the team behind quants.md. Everything on this site — the lessons, tools, practice bank and vault browser — is free and open source, and stays that way whether or not you use AgenKit. You can build any of these strategies by hand; the checklist lesson lists everything you need.

esc

Try , , or .

↑↓ move↵ open