ITP360ITP360
HomeAbout UsServicesSupportInsightsFAQContact
(305) 317-0959Client Portal
ITP Cortex

Private AI for Business.
Your Data Stays Yours.

ITP Cortex is AI your business can actually put real work into. It plugs into the software you already run, it handles material you would never paste into a free chatbot, and it bills on the same ITP360 invoice as everything else. We charge you for what you use. We do not keep what you send.

Schedule Free Consultation(305) 317-0959
HIPAA
Paperwork In Place
One Bill
Not Five Vendors
Miami
People You Can Call

Drop-In Replacement

Change The Address.
Nothing Else.

Cortex works the same way the big AI services do, so software that already talks to one of them will talk to us. Your developer points it at api.itp360.com/v3/cortex and the job is done. No rebuild, no migration project, no retraining your team on a new tool.

Pick a different model whenever you want, even job by job, so the routine work runs somewhere cheap and the hard work runs on the best model available. Ask the same thing twice and the second one costs less.

from openai import OpenAI

client = OpenAI(
    api_key="itp_your_key_here",
    base_url="https://api.itp360.com/v3/cortex",
)

resp = client.chat.completions.create(
    model="itp-balanced",
    messages=[
        {"role": "user", "content": "Summarize this intake note."},
    ],
)

Your Data

We Charge For Usage.
We Do Not Keep Your Content.

Most AI companies hold onto what you send them for a while, and ask you to trust a policy page that they can rewrite whenever they like. We built ours so there is nothing to hold onto. Your questions and the answers pass through and are gone.

What we keep

Enough to put a number on your invoice.

  • how much you used
  • which model you picked
  • what it cost
  • when you used it

What we never keep

Not hidden. Not held for 30 days. Never saved at all.

  • the questions you ask
  • the answers you get back
  • the documents you paste in
  • your patient, client and case records

A privacy promise is only as good as the company making it. We would rather not be in a position to break ours, which is why we are comfortable putting medical and legal work through this.

Model Matrix

Several Labs,
One API Key.

You choose a model by name. When something better or cheaper comes along we move that name onto it, so your upgrade happens without anyone touching your setup.

Model nameBuilt byReach for it when
itp-fastAnthropicClassification, extraction and short replies at volume.
itp-balancedAnthropicThe default for most application work.
itp-maxAnthropicLong reasoning and large documents.
itp-cortex-*Together AI, OpenAILlama 3.3 70B, Qwen, GLM, DeepSeek, Kimi, GPT-5.6 variants and embedding models.

Where It Gets Used

Work You Would Not Send
To A Consumer AI Account.

Documents That Contain PHI

Intake notes, referrals, prior authorizations and clinical summaries. We count the usage. We do not keep the note.

Agents Inside Your Own Software

Drop it into software your team already runs. Most tools only need the address changed, and everything you built keeps working.

Card and Payment Workflows

Dispute summaries, fraud triage and reconciliation notes for businesses carrying PCI DSS obligations on the systems around the model.

Bulk Classification and Extraction

Send the easy work to a fast, cheap model and the hard work to the best one, in the same job, without signing up anywhere new.

Retrieval and Embeddings

Search across your own documents and have the AI answer from them. Same service, same bill, no second vendor.

Voice and Contact Center

The same inference layer our Voice360 platform uses for call summaries and after-hours agents, available to your own applications.

Why Buy It From Us

The Model Is The
Easy Part.

BAA Coverage Already In Place

We hold BAA coverage with the providers behind Cortex, including Anthropic. You sign one agreement with ITP360 rather than negotiating a separate BAA with every lab you want to call.

Spend Caps That Stop Spend

Set a monthly limit. You get a warning as you approach it and a hard stop at it. If something misbehaves at 2am it gets shut off, and you are not reading about it on the next invoice.

Model Names You Control

You pick a model by name. When a better one comes out we move that name onto it, and nothing on your side has to change. Switch models whenever you want.

One Invoice, No Card With A Lab

Usage is metered per call and lands on the same ITP360 invoice as your managed IT, voice and fiber. No separate vendor bill, no per-seat minimum, no credit card on file with a US AI company.

Compared To Rolling Your Own

Two Ways To Buy
The Same Models.

Going direct to each lab works. It also means a key, a contract and a card per vendor, and a compliance conversation you run yourself.

Separate Vendor Accounts

One signup per AI lab

  • A separate signup for every AI company
  • Another bill, on another card
  • You chase the HIPAA paperwork yourself
  • You find out you overspent when the bill lands
  • Streaming responses supported

ITP Cortex

One service, one login

  • One login for every model we offer
  • On the ITP360 invoice you already pay
  • HIPAA paperwork already handled for you
  • Your spending limit is enforced, not suggested
  • Streaming responses (not supported yet)

Plans

Metered Per Call,
Capped Per Plan.

A monthly base plus what you use. Every plan lets you set a limit that warns you on the way up and stops the spending when you get there.

Metered

Pay for what you call

$49/mo
  • Nothing included, pure pay as you go
  • Pay only for what you use
  • Repeat questions cost less
  • Soft and hard spend caps
  • Billed on your ITP360 invoice
Get Started
Most Popular

Standard

For production workloads

$199/mo
  • $150 of usage included each month
  • Go over and you just pay the difference
  • Repeat questions cost less
  • Soft and hard spend caps
  • Usage broken out per project
Get Started

Committed

For steady volume

$999/mo
  • $900 of usage included each month
  • 20% off anything past that
  • Repeat questions cost less
  • Soft and hard spend caps
  • Usage broken out per project
Get Started

Dedicated

Qualifying workloads

Talk to us
  • Dedicated or self-hosted model capacity
  • Available on request after a technical review
  • Same API surface, same invoice
  • Scoped per workload, not an off-the-shelf plan
Get Started
HIPAA
BAA In Place
One
Login For Every Model
Any Time
Switch Models
Miami
Support Team

Frequently Asked Questions

What developers and compliance officers ask us about ITP Cortex.

How much you used, which model you picked, what it cost and when. That is what goes on your invoice and that is all we record. We do not keep the questions you ask, the answers you get back, or any document you put in front of it, and none of it is used to train anything. This is not a setting you have to go turn on. There is nowhere for that material to be saved in the first place.
Almost never. Cortex works the same way the big AI services do, so most software only needs the address and the key changed. Whatever your team already built keeps running. If you are not sure about a specific tool, send it to us and we will tell you before you commit to anything.
Not yet. If you want the answer to appear word by word as it is being written, the way a chat window does, that is still on our build list. Everything else works today. If your app needs it, tell us before you plan a switch.
We sign a Business Associate Agreement with healthcare clients, and the paperwork with the companies whose models sit behind Cortex is already done, including Anthropic. Add the fact that we never keep what you send and the AI side of it is covered. Your own systems still have obligations of their own, your records, your logins, who can see what, and that is work we already do for medical practices here.
A fast one, a balanced one and a heavyweight, plus a wider list including Llama, Qwen, DeepSeek, Kimi and GPT. You pick by name and switch any time. When a better model comes out we move that name onto it, so you get the upgrade without changing anything.
Two limits. The first one warns you. The second one stops it dead. The check happens before the work runs, so once you hit your ceiling nothing more gets spent, no matter what is looping in the background.
On request, for workloads that justify it. Cortex resolves a public model name to an upstream at request time, so adding dedicated or self-hosted capacity is a configuration change on our side rather than a rebuild. Today the live model matrix runs on provider capacity. If you have a workload that needs isolated inference, book a technical review and we will scope it against real volume rather than guesswork.
Same idea, less hassle. Big companies use Azure OpenAI because they want serious AI under a real business agreement instead of a consumer signup. Cortex is that, with one login instead of one per AI company, the HIPAA paperwork already handled, a spending limit that actually holds, and the charge on the invoice you already pay. You also get a phone number with people in Miami on the end of it.

Want a key to test against?

Book 15 minutes. We will talk through your workload, issue a key, and you can have a request working before the call ends.

Book Consultation(305) 317-0959