NEW RESEARCH: Your Sandbox Is Made of Glass

Read

Trinitite

Pricing

Govern your AI in one line of code. Pay a small fee per governed call.

No subscriptions. No per-seat fees. No monthly minimums.

Just like a card network: you keep paying your AI provider directly, and we add a small fee on top of each call we govern. If you make zero calls, you pay zero.

What a governed call costs

Two modes. Choose per call, or set a default.

This is the whole pricing model for the AI you build and run through us.

Watch

monitor

~0.5%

of the provider cost of each call

Observe every call for compliance, policy, drift, and shadow-AI risk. Records everything. Doesn’t intervene.

$10,000/mo of OpenAI → about $50/mo to watch.

Protect

guardian

~2%

of the provider cost of each call

Actively blocks, masks, or auto-corrects unsafe calls inline, before they cause harm. Masking, human-review routing, and secure agent identity are included.

$10,000/mo of OpenAI → about $200/mo to protect.

Estimate your bill

no seats · no minimums

Your monthly AI provider spend

$14,000/mo

Watch · ~0.5%

$70/mo

Observe every call for compliance, policy, drift, and shadow-AI risk. Records everything.

Protect · ~2%

$280/mo

Block, mask, or auto-correct unsafe calls inline, before they cause harm.

That's the whole bill. You keep paying your AI provider directly; we add a small fee only on the calls we govern.

Everything else you can meter

Pay per unit. Nothing between uses.

Browser & desktop AI

Shadow AI: ChatGPT, Claude, Gemini, Copilot

Watch

from $0.001 per governed exchange

Protect

from $0.005 per governed exchange

Plugs into your existing web gateway, so no software on employee laptops. Volume brings the per-exchange price down.

Build & improve your models

Fine-tune, distill, or generate training data

GPU time

$12 per GPU-hour ($0.20/min)

Typical run

~90 min ≈ $18

Billed by the minute you use. Nothing when you’re not.

Testing, evals & red-teaming

Test your agents before you ship

Eval / test scenario

from $0.01 per scenario

Red-team probe

from $0.02 per adversarial probe

A 50-scenario red-team ≈ $1 to $2. Pay per test, nothing between runs.

Compliance documents

Signed, audit-ready, on demand

Model card / factsheet

flat per document

Data-subject-rights export

flat per export

EU AI Act, ISO 42005, NIST, SOC 2, generated from data you already have.

Always free

A whole observability and GRC stack, included free.

You never pay for access, seats, or looking at your own data. The tools most teams buy separately are just part of governing a call with us.

✓

Full AI observability

Live dashboards, logs, request traces, and session replay on every governed call. No separate observability bill.

✓

Spend & FinOps

Spend tracking, per-team and per-agent chargeback, a model catalog, and smart routing across providers.

✓

Compliance posture

Framework crosswalks (SOC 2, EU AI Act, and more), your risk register, and your live AI risk score.

✓

The full audit trail

Every decision, in order, queryable and exportable. Hand an auditor a report instead of staffing a project.

✓

Unlimited team members

No per-seat fees, ever. Add everyone who needs to see the data, from engineering to the board.

✓

Test mode

Build and test the entire integration free, with no live calls and no billing, before you point it at real traffic.

✓

Deterministic decisions

The same call always resolves the same way, so what you test is exactly what you ship.

✓

Model catalog & routing

Compare providers, get recommendations, and route each call to the right model automatically.

Replaces a stack of paid tools

LLM observability

Session replay

Cost & FinOps dashboards

GRC posture & crosswalks

Model gateway & router

Risk register

No extra contracts, no per-seat add-ons. It is all included the moment you govern your first call.

How billing works

Simple, and only ever for what you use.

You pay your AI provider directly.

We never resell their tokens. Our fee is a small percentage on top, only on the calls we govern.

The percentage is the price.

For normal calls, that’s all there is: about 0.5% to watch, about 2% to protect.

You only pay for what you use.

No call, no charge. No seats, no monthly fee, no minimum.

One usage invoice.

Everything metered rolls up into a single, itemized bill.

Very large inputs have a tiny minimum based on their size, because a huge document costs us more to read. Everyday calls never hit it.

“What will this cost me?”

Three real examples.

Solo developer / small app

100,000 protected calls/mo, ~2¢ each (~$2,000 provider spend)

~$40/mo

Growing startup

5M calls/mo, mostly watched with key flows protected (~$50,000 provider spend)

~$500/mo

Enterprise

Large fleet, full protection, browser-AI governance, air-gap

Volume rates + add-ons

Add as many teammates as you want, free. High volume earns lower rates on the same per-call model, so the percentage just gets smaller as you grow.

Pricing FAQ

Answered.

How does Trinitite pricing work?

Trinitite is priced like a card network: you pay a small fee per governed AI call and nothing else. No subscriptions, no per-seat fees, no monthly minimums. Watching a call costs about 0.5% of the provider cost of that call; actively protecting it (block, mask, or auto-correct inline) costs about 2%. You keep paying your AI provider directly, and we add a small fee on top only on the calls we govern. If you make zero calls, you pay zero.

What does governing browser and desktop AI cost?

For AI your team uses that you did not build, such as ChatGPT, Claude, Gemini, Notion AI, or Copilot, there is no provider bill for us to take a percentage of, so it is a flat, tiny fee per governed exchange (one full back-and-forth). Watching starts at $0.001 per exchange and protecting starts at $0.005 per exchange. It plugs into your existing web gateway, so there is no software on employee laptops, and volume brings the per-exchange price down.

Do I have to pay per seat?

No. Team members are always free: unlimited seats, no per-seat fees, ever. Dashboards, logs, the full audit trail, compliance posture and framework crosswalks, your AI risk score, spend tracking, and session replay are all included at no charge. You can build and test the entire integration for free in test mode before you point it at real traffic.

How does pricing change at scale?

The percentages are standard list pricing. High volume earns lower rates on the same per-call model: you never switch to a different pricing shape, the percentage just gets smaller as you grow. Optional deployment add-ons for regulated or large orgs (air-gapped/on-premise install, org-wide browser-AI governance, identity federation with your IdP, mission-critical 24/7 support) are scoped to your environment.