ComplyHat

ComplyHat is a software platform designed to automate compliance documentation and testing for AI models across multiple regulatory frameworks. It enables organizations to connect AI agents such as Claude or Codex and generate compliance reports based on real tests conducted on their own models. These reports are mapped to frameworks including the EU AI Act, SR 26-2 (Joint Fed/FDIC/OCC model-risk-management guidance), NIST AI RMF, and ISO/IEC 42001, and are exported in regulator-ready DOCX format. The platform is intended for compliance officers, ML engineers, fintech compliance teams, AI product teams, and enterprise deployers in sectors such as banking and insurance.

ComplyHat operates through a remote MCP (Model Compliance Proxy) URL, which users configure in their AI agent host. Supported hosts include Claude Code, Claude Desktop, Codex Desktop, Codex CLI, OpenClaw, and NemoClaw, as well as any MCP-capable harness like LangGraph, CrewAI, or AutoGen. Integration is completed via OAuth, and there is no need for CLI installation or plugin downloads. The platform does not access users' source code, model weights, or training data; instead, only the necessary evidence for compliance tests is passed through, with computations performed in memory and only summary artifacts persisted. This approach is designed to streamline security reviews and protect sensitive data.

The system computes compliance metrics directly on users' model data and checks them against framework-specific thresholds. Every approved compliance report is hash-locked and stored in an immutable log, recording who signed off and when, providing a tamper-proof audit trail. Reports and compliance decisions are accessible through a dashboard, and each filing builds on previously approved positions. Human approval is required for all compliance decisions, with the software explicitly stating that it does not provide legal advice; users and their counsel are responsible for final verification before regulator submission.

ComplyHat is offered under three pricing tiers: Solo (one seat, all frameworks, unlimited AI models, audit trail, with a 14-day free trial), Team (up to five seats, higher quotas, team-shared compliance wiki, priority support), and Enterprise (custom seat count, dedicated support, volume pricing, custom data residency, and founder-direct onboarding). Billing is monthly or annual, with savings available for annual plans. The platform charges per MCP install rather than per AI model, and all sales are final.