Turn the AI you already use into your compliance team
Connect to Claude or Codex, ask for an EU AI Act or SR 26-2 report, and your assistant drafts it from real tests on your own model. You approve it, and ComplyHat saves the audit-ready document and remembers every decision for your next filing.
Four framework templates. The documentation regulators expect.
SR 26-2
Joint Fed/FDIC/OCC model-risk-management guidance for large bank holding companies (effective 2026-04-17).
EU AI Act
Regulation (EU) 2024/1689. Annex IV technical-file sections required by Article 17 (deadline 2 August 2026).
NIST AI RMF
NIST AI 100-1. Govern, Map, Measure, Manage attestations. Increasingly cited as a state AI law safe harbor.
ISO/IEC 42001
ISO/IEC 42001:2023 AI Management System. Annex A controls and management-review artefacts.
Claude can describe compliance. ComplyHat makes it hold up.
The compliance work your team hands to consultants, now delivered as software through the AI you already use. You bring the judgment. ComplyHat brings the math, the memory, and the paper trail, and never touches your code or data.
Real numbers, not descriptions
Claude explains the metric. ComplyHat computes it on your model and checks it against the threshold, every time.
Evidence, not a transcript
Every approved report is hash-locked in an immutable log: who signed off, and when.
Your regulator's format, not a copy-paste
Mapped to all four frameworks and exported as clean, regulator-ready DOCX.
Memory that compounds
Connect in 30 seconds. Every filing builds on the positions you already approved.
Not a chat window. The dashboard your examiner opens.
Every report your assistant files lands here as a read-only record: the same evidence, hash, and DOCX an auditor expects to see.
Real tests, not prose
Scores computed on your actual model data and checked against each framework's threshold, the same way every time.
Tamper-proof audit trail
Every approved document is fingerprinted with a sha256 hash and written to an immutable log with who approved it and when.
Regulator-ready DOCX
Download the exact version you signed off, formatted the way your examiner expects, not copied out of a chat window.
Plans for Every Stage
Pay per MCP install, not per AI model. Start with a 14-day free trial of Solo or Team. Enterprise is custom-priced.
SOLO
$499/mo
- 14-day free trial
- 1 seat
- All 4 frameworks
- Unlimited AI models
- MCP audit trail
Start Free Trial
For solo compliance officers and individual ML engineers
TEAM
$2,499/mo
- Everything in Solo
- Up to 5 seats
- Higher monthly quotas
- Team-shared compliance wiki
- Priority support
Start Free Trial
For fintech compliance teams and AI product teams
ENTERPRISE
Custom
- Custom seat count
- Dedicated support
- Volume pricing
- Custom data residency
- Founder-direct onboarding
Request a Demo
For mid-tier banks, large insurers, and enterprise deployers
All sales final. No refunds. See our Terms.
Frequently Asked Questions
Everything you need to know about ComplyHat and AI compliance documentation.
How does ComplyHat handle multiple regulatory frameworks?
The engine is framework-agnostic: you pick a framework per run and ComplyHat applies its thresholds. Run the same test against any of the four (SR 26-2, EU AI Act, NIST AI RMF, ISO/IEC 42001). Only the thresholds and the rendered obligations change.
What counts as a seat?
One human who has authenticated to ComplyHat through an MCP host. One person equals one seat, no matter how many devices or hosts they use. AI models are uncapped on every plan. You only pay for the humans using the MCP.
How do I get started?
Sign up, connect via OAuth, and the MCP client registers itself with your agent. We have first-class integrations with Claude Code, Claude Desktop, Codex Desktop, Codex CLI, OpenClaw, and NemoClaw, all sharing the same MCP URL. Any other MCP-capable harness works the same way.
Which agent frameworks does ComplyHat support?
Six first-class hosts via one remote MCP: Claude Code, Claude Desktop, Codex Desktop, Codex CLI, OpenClaw, and NemoClaw. Any other MCP-capable harness (LangGraph, CrewAI, AutoGen) works through the same endpoint.
How do I install ComplyHat?
Paste the remote MCP URL (https://complyhat.ai/api/mcp) into your host config and authorize via OAuth. No CLI install, no plugin download. The quickstart at docs.complyhat.ai has the exact snippet for each host.
Does ComplyHat need access to our code, models, or training data?
No, there is nothing to grant access to. ComplyHat never reads your source code, model weights, or training data. Your host passes only the evidence a test needs. ComplyHat computes in memory and persists only the summary artifacts, never raw data. That keeps your security review short.
Is my data secure?
Yes. Encryption at rest and in transit, row level tenant isolation, immutable audit trails, and human in the loop approval for all compliance decisions. Agents never mark a model compliant. Humans sign off.
Is ComplyHat legal advice?
No. ComplyHat renders structured compliance documents from your test results. It is not a law firm and gives no legal advice. Every output is a draft. You and your counsel verify it and own the decision before anything reaches a regulator. See our Terms.