System Overview

    Kronos AI Trader is an AI-powered trading decision support and execution system. This documentation provides technical details about the platform architecture and capabilities.

    High-Level Architecture

    Frontend

    • • React 18 with TypeScript
    • • TailwindCSS + Shadcn/UI
    • • Real-time WebSocket data
    • • React Query state management

    Backend

    • • PostgreSQL with RLS
    • • Edge Functions (Deno)
    • • AES-GCM encryption
    • • Rate limiting & audit logs

    Data Flow

    Exchange API
    WebSocket Stream
    Analysis Engine
    Decision Layer
    Execution

    About the Kronos AI developer documentation

    This reference documents how to integrate with Kronos AI Trader programmatically: authenticating requests, connecting an exchange account, reading bot state, retrieving trade history and inspecting the risk parameters applied to each execution. It is written for developers and quantitative teams who want to embed Kronos AI data into their own dashboards, reporting stacks or compliance workflows.

    Every endpoint is scoped to the authenticated account. Credentials are transmitted as bearer tokens over TLS, exchange API keys are stored encrypted at rest and no endpoint exposes another account’s positions, balances or configuration. Rate limits protect the shared execution path; clients that exceed them receive an explicit error rather than partially processed data, so integrations should retry with exponential backoff.

    Configuration values described here — stop loss, take profit, leverage, position size, minimum confidence and strategy selection — are the same values the execution engine reads at runtime. There is a single source of truth: changing a parameter through the API changes the behaviour of the live bot immediately, and the change is recorded in the audit trail together with the account that made it.

    The documentation reflects the current production behaviour of the platform. Breaking changes are announced before deployment and older response shapes remain readable for a transition period. If a response does not match what is described here, report it through the support channel with the request identifier so it can be traced against server logs.