Project

AI Trading Audit Platform

A production-grade audit and orchestration stack for AI-assisted trading recommendations. It enforces deterministic policy gates, emits append-only audit events (with optional hash chaining), and propagates trace IDs for end-to-end compliance review.

Orchestrator enforces trust boundaries and policy gates.
Audit MCP stores immutable events with optional hash chaining.
Trace IDs propagate end-to-end for auditability.
Risk/compliance checks are hard gates, advisory is optional.

System overview

Audit flow

Trust boundary

Key capabilities

  • Append-only audit logging with optional tamper-evident hash chain.
  • End-to-end trace propagation for compliance review.
  • Strict separation between policy enforcement and advisory output.
  • Fail-closed behavior when audit logging is unavailable.

Core services

  • Orchestrator (FastAPI + Pydantic) for request validation and policy gates.
  • Audit MCP (FastAPI + SQLite) for immutable event storage.
  • Shared schemas for trade requests, responses, and audit events.