New  — Introducing TrustLog Guard, the financial immune system for AI agents
Live demonstration — AWS EC2

AI agents spend money.
Nobody's watching.

Comptex builds financial governance for autonomous AI compute. We monitor, benchmark, and circuit-break AI spending before it becomes a loss you can't explain.

£5.27
Burned in 47 seconds
without governance
$1.43
Caught by
circuit breaker
73%
Capital
preserved
47s
Time to total loss
without governance

Live experiment. Real API costs. March 2026.


Experiment

The 47-Second Flash Crash

Same AI agent. Same task. Same recursive loop. One run had no financial governance. The other had TrustLog Guard active.

Without governance — £5.27 lost

UNHEDGED · 47 seconds · Manual kill required · £403/hr projected burn

With TrustLog Guard — $3.84 saved

HEDGED · Circuit breaker at $1.43 · Auto post-mortem generated

auto post-mortem — /risk
TrustLog Guard — Risk Assessment RISK: 7/10 WHY: ▸ Cost-at-Risk breached +2 Spend exceeds 95th percentile ▸ Model concentration critical +2 HHI = 0.82 (86% on opus) ▸ Burn rate spike +1 8.3x above normal ▸ Budget at 84% +1 Daily limit approaching
COST-AT-RISK CaR(95%): $14.20 Today: $14.43 BREACH
CONCENTRATION HHI: 0.82 HIGH If opus pricing +20% → costs +17%
→ Loop halted. Move recursive tasks to haiku. Saves $1.80/day. Drops HHI to 0.45.

In traditional finance, every traded asset has governance: risk limits, position monitoring, circuit breakers. AI compute has none. We are building the governance layer that autonomous compute requires.


Market

The problem is real and growing

We didn't invent this problem. These founders are living it.

Saved $25,000 per month by switching one AI model. Margins went from 50% to 80% overnight.
Pieter Levels — Photo AI
AI model costs swung from $0.003 to $0.15 per image — 50x — then back to $0.05.
Pieter Levels — on cost volatility
Running 14 autonomous AI agents for marketing. Every 15 minutes they check a shared dashboard.
Brett Goldstein — agent orchestration
46.6% of US businesses now pay for AI. Average contract heading to $1M in 2026.
Ramp AI Index

Products

Two altitudes. One thesis.

AI spending is invisible and it shouldn't be.

Agent Level

TrustLog Guard

Financial immune system for AI agents. Passive detection, circuit breakers, risk scoring. Install it. Forget about it. It watches.

GitHub →
Business Level

NetMRR

Your real MRR after AI costs. Which customers are profitable. Which aren't. The number Stripe doesn't show.

netmrr.com →

Collaborate

This field doesn't exist yet.
Help us define it.

Financial governance for AI agents has no papers, no benchmarks, no standards. We're looking for people who want to build the foundations.

Research

Academic partners to co-author foundational work. We have implementations, data, and a framework ready to formalise.

Technical

Engineers in the AI agent ecosystem. If you build agents, orchestrators, or infra — we're building the financial layer underneath.

Institutional

Universities, accelerators, research bodies. Open to fellowships, joint programmes, and sponsored research into compute risk.

If this resonates, we'd like to hear from you.


People
AB

Anouar Boussif

Quant-finance technologist engineering deterministic risk management for autonomous systems. Building the financial liquidity layer for the agentic economy.

X · GitHub · netmrr.com · Direct