Glossary
This page defines the terms you’ll see across the rest of the documentation. Each entry links to the page where the concept is explained in depth, so other pages can reference vocabulary without redefining it. Entries are alphabetical.
ACP — Agent Client Protocol. A transport where Cortiq drives a local agent (for example claude-agent-acp or codex-acp) over JSON-RPC instead of calling a REST API. The agent runs under your own local subscription. See AI Providers.
Agent permission / capability grant — A persisted grant that lets the agent take a specific class of non-trade action (for example a web action) without prompting each time. Granted once from an actionable System Message, then auto-allowed. Trade execution is never covered by a capability grant — it stays on its own approval gate. See Agent permissions.
Autonomous session — A session whose trading loop is driven by Cortiq’s internal workflow engine. The default mode for most users. See Sessions.
Backtest — A historical M1 replay of a session or playbook against past market data. Backtesting is the only license-free path in Cortiq. See Backtesting.
Copilot — The in-app Copilot dock, opened with Ctrl+K. It authors and edits entities (playbooks, profiles, preparation notes, and more) on your behalf, behind the same approval gate as the agent. See Cortiq Copilot.
Data package — The configurable bundle of market data, indicator values, screenshots, account state, and supporting context that Cortiq sends to the AI on each cycle. See Playbooks & Data Packages.
Day start time — The user-configured local time at which Cortiq’s daily P/L and trade-count counters reset. Set in risk settings, not in the OS clock. See Risk Management.
External MCP — A transport where an outside MCP-compatible client (for example Claude Desktop) drives the session loop by calling Cortiq’s MCP tools and posting a decision back. Cortiq itself never calls an LLM in this mode. See MCP and Agent Integration.
External MCP session — A session whose trading loop is driven by an external MCP client via the External MCP transport. The internal workflow engine is not running. See MCP and Agent Integration.
Exploration run — A watch-and-learn research run that steps an AI through historical candles of one instrument with no playbook and no positions, and writes what it notices to the knowledge ledger. Its instance-finder script runs in a network-denied sandbox. See Research.
Instrument Hub — The per-symbol command center at /instruments/{symbol} that aggregates every playbook, session, backtest, research run, profile, and knowledge entry for one market, headed by the honest ★ mastery rating. See Instrument Hub.
Instrument profile — A reusable per-symbol context document (typical session, common ranges, behavioral notes). See Trading Cycle: Instrument Profiles.
Integration mode — How Cortiq reaches the provider: API, ACP, CLI, or External MCP. Each provider supports one or more. See AI Providers.
Knowledge ledger — The append-only, instrument-scoped evidence trail on the Instrument Hub. Each entry is an Observation (a single noticed instance) or a Learning (a lesson seen across several instances), stamped with its source and a confidence weight. See Instrument Hub.
Lesson — A durable, instrument-scoped takeaway produced by a Session Review and carried forward into later sessions on that instrument. See Session Review.
Local model — A self-hosted, OpenAI-compatible model served locally through Ollama or LM Studio. Reached over the API transport against a local endpoint. See AI Providers.
Managed data file — A reusable, pre-built parquet dataset for one instrument — a symbol, timeframe, chosen columns, indicators, and an editable build script — that a research or exploration run reads instead of authoring data each time. Managed on the Research page’s Data Files tab. See Research.
Mastery rating (★) — The Instrument Hub headline. It equals the number of playbooks promoted to Production maturity for a market, and nothing else moves it. See Instrument Hub.
MT5 bridge — The ZeroMQ connection between Cortiq and a running MetaTrader 5 terminal on the same machine. See MetaTrader 5 Integration.
OpenRouter — An OpenAI-compatible provider Cortiq reaches over the API transport; the endpoint can be repointed at any other OpenAI-compatible service. See AI Providers.
Playbook — A reusable strategy definition: the rules, filters, and operating logic the AI follows during a session. See Playbooks & Data Packages.
Preparation package — A reusable context layer (for example, a daily bias note) that supplements the data package on every cycle. See Trading Cycle: Preparation Packages.
Provider — An AI service Cortiq talks to: ChatGPT (OpenAI), Gemini (Google), Claude (Anthropic), OpenRouter / OpenAI-compatible, or a local model. You bring your own. See AI Providers.
Research workspace — A single-instrument idea lab. Each workspace is scoped to one symbol and holds research runs and their artifacts; output can seed a playbook, instrument profile, or preparation notes. See Research.
Risk limit (global) — A risk rule that applies across all MT5 accounts combined: total daily drawdown, total weekly drawdown, total daily trades. See Risk Management.
Risk limit (per-account) — A risk rule that applies to one MT5 account only. Independent of other accounts. See Risk Management.
Sentiment report — A reusable context document summarizing market positioning or news sentiment, attached to a session for richer AI context. See Trading Cycle: Sentiment Reports.
Session — A trading session. The smallest unit of repeatable Cortiq operation: an MT5 account, one symbol, a provider, a time window, and risk settings, executed against a playbook. Each session is locked to a single instrument. See Sessions.
Session Review — A one-click AI review of a finished session that scores what happened and produces suggestions plus durable instrument lessons. See Session Review.
Skill — A reusable Markdown instruction template you write once and inject into generation flows. See Skills.
System Message — An in-app notification in the System Messages inbox. Some are informational; some are actionable permission cards used to grant a capability. See System messages.
Trade idea — A standalone trade thesis captured outside the playbook library, useful for one-off opportunities. See Trading Cycle: Trade Ideas.
Virtual mode — A session mode that runs forward on live data but holds back the order instead of sending it to MT5. Useful for evaluation, rehearsal, and challenge environments. Distinct from a Backtest, which replays history. See Execution Modes & Notifications.