AI decisions your board can defend. With the evidence attached.
Quantamix sells TraceGov.ai, a hosted evidence platform run from AWS Frankfurt that records, for every AI output, the policy, data, model version and approval behind it, and a fixed-scope two-week engagement that puts it under one of your AI systems. You keep the records. Every record can be checked by someone who was never in the room.
One AI decision with its policy, data, model version, approval and outcome hashed into a Merkle root, signed, anchored, and checked by a verifier with no access to the infrastructure.
Source: Record design: GraQle loop article, section 7. Hashes illustrative.
The market in 2026
Agents are in daily use. The evidence behind their decisions is not.
Independent surveys published in 2026. Each number links to its source.
The system in six pictures
What we build, drawn from the code and the papers
The record
One decision with its evidence hashed, signed, anchored and checkable by an outsider.
The suite
Four products, three services, one knowledge graph.
The gates
Seven hard gates; a single FAIL stops the action.
The loop
ANCHOR to COMMIT, run through a real 146-node project graph.
The argument
Seven papers and a board report as one argument, every DOI shown.
How we read the market
56 signals, 24 theses, 9 proposed capabilities.
Products
The governance suite
Four products on one knowledge graph: scoring, reasoning, retrieval and adoption, each with its own record of why it answered the way it did.
TraceGov.ai
EU AI Act alignment scored across the five TRACE dimensions
Hosted platform, AWS Frankfurt (eu-central-1)
- Graph-based retrieval over the regulatory text
- Patent application EP26162901.8 (pending)
- Cryptographic audit trail, 7 year retention
GraQle
The intelligence engine
SDK on PyPI, v0.85.0, source-available, 13 model backends
- Multi-agent cross-verification
- Zero LLM calls at retrieval
- Source-available: 2,009 tests
TAMR+
Trust-aware multi-signal document retrieval
TAMR+ Lite on PyPI (Apache 2.0); working paper on SSRN and Zenodo
- Patent application EP26162901.8 (pending)
- TAMR+ Lite: standalone Python package on PyPI, Apache 2.0
- Working paper on SSRN and Zenodo
FrictionMelt
AI friction intelligence
Diagnostic framework: 95 friction points across 8 layers
- 95 friction points across 8 layers
- Compound failure trigger identification
- Prioritised by business impact
Services
Three ways to start
Each one tells you what you get, how long it takes and how it is priced before you commit to anything.
Free AI Quick Scan
- What you get
- 30 governance questions across AI inventory, EU AI Act and GDPR. An instant maturity score and risk flags, readable at board level.
- How long
- 15 to 30 minutes
- How it is priced
- Free. No sign-up.
Diagnostic briefing
- What you get
- Your use case and your regulatory exposure, worked through in nine structured steps. No sales deck. You bring one AI decision; you leave with its evidence chain and a 90-day roadmap.
- How long
- 45 minutes
- How it is priced
- Free. No commitment.
Two-week proof of concept
- What you get
- One AI system of yours. A full TRACE score report at the end, with the six-hour board report, so the gap is shown in a form your board can read. A redacted sample of both reports is available on request.
- How long
- Two weeks
- How it is priced
- Fixed scope, priced after the briefing.
Case study
Philips Personal Health
Content for 500+ SKUs, produced on CrawlQ with brand rules checked before publication. The numbers below are the ones Philips verified.
Philips Personal Health
Healthcare technology
“CrawlQ's AI platform shifted my perspective on what AI can offer at enterprise level.”
Founder
Built by someone who has had to defend the numbers
Harish Kumar
Founder, Quantamix Solutions B.V.
Harish brings 22 years across quantitative risk and AI in regulated industries, most of them spent producing numbers that a regulator or an auditor would later check. That record is behind the three European patent applications pending on the retrieval and reasoning methods inside the governance suite.
Bring one AI decision. Leave with its evidence chain.
Pick a decision your organisation already makes with AI and we will show you, step by step, what a regulator could ask for and what you could hand over today.
No sales deck. 45 minutes. Free. You leave with the evidence chain for one decision.