Trading cycle overview

This page explains the architecture of one Cortiq cycle. By the end you’ll know which pieces feed each phase and why the session — not the AI — is the right object to think outward from.

What this is

The clearest way to understand Cortiq is to think in terms of one session running one repeatable trading cycle. That cycle is more than an AI prompt — it’s a controlled operating loop assembled from a session, one or more playbooks, one data package, optional support layers, the live MT5 state, and the trade history.

Each cycle moves through six phases. Understanding the phases makes it obvious which Cortiq object you should change to influence which behavior: a problem in the entry quality is rarely a “the AI is bad” problem; it’s almost always an issue in the data package or playbook feeding that AI.

How it fits into Cortiq

flowchart LR
  Pre(Preparation) --> Gather(Data gathering)
  Gather --> Decide(AI decision)
  Decide --> Risk{Risk check}
  Risk -->|approved| Exec(Execution)
  Risk -->|rejected| Skip(Skip + log)
  Exec --> Manage(Trade management)
  Manage --> Review(Review + journal)
  Skip --> Review

The trading cycle has six phases. Preparation and review bookend the cycle; the AI decision is one phase among many, not the whole loop.

The session sits in the middle of every phase: it decides which account, symbol, provider, playbook, and data package are active, and when the session is allowed to run.

How to use it

Read the cycle from the session outward

Start by understanding what the session controls, then drill into each phase as needed.

Phase	Driven by	Read more
Preparation	Preparation packages, instrument profiles, sentiment reports	Supporting context
Data gathering	Data package, MT5 bridge	Data package design
AI decision	Playbook, AI provider	Playbook design
Risk check	Risk validators (global + per-account)	Risk management
Execution	MT5 integration	MetaTrader 5 integration
Trade management	Playbook management rules	Session trades and timeline
Review	Journal, analytics, conversations	Journal & analytics

Add support layers when they earn their place

Playbooks and data packages cover the core. Support layers — session instructions, preparation packages, instrument profiles, sentiment reports — let you add structure the AI shouldn’t have to rediscover every cycle.

Add a layer when its absence would noticeably hurt cycle quality. Don’t add layers reflexively; a noisy preparation package is as bad as a noisy data package.

Reference

How the pieces fit together

Layer	Role	Typical question it answers
Session	Operating container	What is this workflow allowed to do?
Playbooks	Strategy rules	What setups are valid?
Data package	Live market payload	What does the AI see every cycle?
Trade ideas	Specific theses	What discretionary opportunities are tracked right now?
Preparation package	Cached analysis	What slower-moving structure should already be known?
Instrument profile	Long-lived behavior	How does this instrument typically behave?
Sentiment report	News and macro	What external pressure matters right now?
Session trades and timeline	Execution record	What happened and why?

Where most quality problems come from

Symptom	Most likely cause
AI improvises too much	Playbook is too vague.
AI is paralyzed by noise	Data package has too many timeframes or indicators.
Decisions feel disconnected from market context	Missing support layer (preparation, instrument profile, sentiment).
Same trade keeps recurring incorrectly	Invalidation logic in the playbook is missing or weak.