Recorded on a stock laptop · network cable unplugged
4.8 · hundreds of enterprise customers
AirgapAI · With Blockify™ Technology Available

AirgapAI Chat

The enterprise AI assistant that never sends a byte off your laptop

The Experience

Open the app, click one of 2,800+ ready-made workflows, drop in your document, and ship expert output in seconds. Then pull the network cable mid-answer and watch it keep typing — the model, the workflows, and your data all live on your machine.

Most Popular
Monthly — free for 7 days
Full Pro access · $0 today · no card required
$1.15 / day
$35 /month after trial
Best Value
Perpetual License66% less than cloud
Own it forever · free updates · data ingestion + workflow builder
$0.64 / day over 3 yrs
$2,060 cloud 3-yr cost¹ $697 one-time
Max Accuracy
Enterprise + Blockify
78× fewer hallucinations on your data · white-glove install · priority support
$1.37 / day over 3 yrs
$1,499 one-time
Start Free — No Card

No charge for 7 days. After your trial, $35/month. Cancel in one click before day 7.

Buying for a fleet, or need a PO / volume quote? Talk to sales
30-day money-back guarantee
Free 7-day trial — no card required
Cancel anytime, inside the app
How do you deploy it?
One laptop: download, install, and get your first answer in about 10 minutes. A fleet: push it to 10,000+ endpoints in an afternoon with Microsoft Intune, JAMF, or SCCM — with role-based workflow libraries and full audit controls.
Will it run on my laptop?
Windows 10/11 with an Intel Core Ultra CPU or any dedicated NVIDIA/AMD GPU, or any Mac with Apple M1 or newer (8GB+ memory). See full system requirements.
What's included
  • AirgapAI Chat application (Windows + macOS)
  • 2,800+ pre-built, click-to-run workflows across 50+ categories
  • 6+ open-source local models (Arcee 4.5B, LLAMA 3.2, Gemma 9B, QWEN, more)
  • Entourage Mode — multiple AI personas in one chat
  • Local document ingestion + custom workflow builder (Perpetual and up)
  • Email support · lifetime updates on perpetual licenses
Trusted by Fortune 500 Companies, the DoD, and Federal Agencies
Government Acquisitions

What is AirgapAI?

AirgapAI is a 100% local AI assistant from Iternal Technologies that runs entirely on your own Windows or macOS laptop — no cloud, no internet connection, and zero data leaving the device. It includes 2,800+ pre-built workflows, 6+ open-source models (LLAMA 3.2, Arcee, Gemma, QWEN), and an optional Blockify data pipeline benchmarked at a 78x aggregate accuracy improvement on enterprise data. Pricing is $35/month after a 7-day free trial, or $697 once for a perpetual license.

How does AirgapAI deliver air-gapped AI?

AirgapAI runs the model, the retrieval and your documents on hardware you control, with no network connection — no cloud APIs, no telemetry, zero data leaving the device. It ships as a turnkey assistant: local LLMs, 2,800+ ready-to-run workflows, and Blockify retrieval with 78x fewer hallucinations, on a standard business laptop. For the underlying idea, see the air-gapped AI concept.

Where does AirgapAI actually run?

AirgapAI is a private AI assistant that runs the whole model on the laptop or server you own. Prompts, documents, chat history and the 2,800+ workflows stay on that machine, with no account sync and no telemetry. Pull the network cable mid-answer and it keeps generating. See what stays inside the boundary.

Run your own numbers: the air-gapped AI for government calculator, the cloud AI token cost comparator, or the cloud AI security assessment.

At A Glance
~3×1
Cheaper than cloud AI over 3 years
89%2
Employee adoption rate
2,800+
Done-for-you workflows included
0
Bytes sent to any cloud
¹ Forrester: $2,060+ loaded 3-year cost per cloud Copilot seat vs $697 once  ·  ² AirgapAI enterprise deployment data vs 20–30% industry-typical adoption
Why We Built It
“Most ‘enterprise AI’ is a data pipeline to someone else's cloud with a chat window on top. We built the version where the intelligence comes to your data — not the other way around.”
— The Iternal Technologies founding team
The Problem

Every prompt your team sends to cloud AI leaves your building.

Contracts, financials, patient records, deal data — piped to someone else's data center, retained under their terms, and useless to you the moment the connection drops. Legal won't approve it. Security won't trust it. And you pay rent on it forever.

Two enclosures drawn side by side: on the left a building inside a perimeter ring that is broken open, with a thread leaving through the gap and running out to a distant tower standing alone; on the right an identical building inside an unbroken ring, where the same thread loops once inside the ring and returns into the building, and nothing at all stands outside it.
Standard cloud AI
Where your prompt goes
01
Your laptop
Prompt + attached documents composed
Egress
02
Your network edge
Data crosses the corporate boundary
Exposed
03
Provider cloud
Processed + retained on third-party servers
Retained
04
Subpoena / breach / training surface
Out of your control from here on
Risk
AirgapAI
Where your prompt goes
01
Your laptop
Model, workflows, data, and answers — all here
Local
02
—
Nothing crosses the boundary
0 bytes
03
—
No third-party cloud exists in this architecture
N/A
04
—
Nothing to subpoena, breach, or train on
N/A
Fig. 01 · Data exposure path · Typical SaaS AI architecture vs on-device inference
01 · On-device processing
100%
Inference, retrieval, and storage all happen on your hardware.
02 · 3-year cost vs cloud
−66%
$697 once vs $2,060+ loaded 3-year cloud Copilot cost (Forrester).
03 · Workflows preloaded
2,800+
Click-to-run tasks engineered by domain experts. No prompting skill needed.
The Solution

AirgapAI runs the entire model on the laptop you already own.

No cloud round-trip. No per-token meter. No data processing agreement to negotiate. The same class of open models the big clouds run — installed like any other desktop app, deployable with the MDM you already use.

AirgapAI's core idea: a sealed instrument cabinet standing closed on a desk, its unbroken front face lit from within by a single cyan seam, while the wall sockets behind it sit empty and its only cable lies coiled and disconnected on the desk — everything that makes it work is inside, and it is joined to nothing.
Results

Measured outcomes. Not marketing.

Every number below has a named source you can check — the same standard we hold our workflows to.

~3×
Lower 3-Year Cost
$697 once vs $2,060+ loaded cost per cloud Copilot seat over 3 years. Deploy to 100% of staff for less than cloud covers 30%.
Forrester TEI, cloud Copilot
89%
Employee Adoption
Because every workflow is a button, not a blank prompt box. Industry-typical AI adoption stalls at 20–30%.
AirgapAI deployment data
$650M+
Pipeline Generated
Fortune 500 sales teams cut proposal cycles from 6 weeks to 2 days and generated $650M+ in net-new pipeline.
Dell Technologies case study
The Architecture

Two technologies. One assistant.

Two components seated in one shared housing: a sealed navy vessel lit by a single cyan seam (the model running on your device) beside an open rack of fitted, refined blocks (Blockify), joined at the base so the pair reads as one instrument.
On-Device Inference
Sensitive data leaving your network per prompt
Cloud AI subscriptionsEvery prompt
AirgapAI0 bytes

The model runs where the data lives. That's why AirgapAI is deployed in SCIFs and satisfies HIPAA, CMMC/NIST, FedRAMP, and GDPR residency by construction.

Architecture: 100% local processing
Blockify™ Data Refinement
Answer accuracy on your own documents
Standard RAG on raw files1× baseline
AirgapAI + Blockify78×

Blockify distills your messy SharePoint, PDFs, and spreadsheets into validated IdeaBlocks before the model can touch them — 78× fewer hallucinations on enterprise data.

Blockify Performance Analysis — 78x methodology
Put Simply

How it works

A console whose face is a grid of identical blank keys: one key is pressed and lit, a finished sheet slides out of the slot beneath it, and the blank writing tablet beside the console lies closed and unused — the work is a button, not a prompt.
01

Install it anywhere — in minutes, not quarters.

Download and run on one laptop in about 10 minutes, or push to 10,000 endpoints in an afternoon with Intune, JAMF, or SCCM. No cloud tenancy, no data processing agreement, no six-month security review — there's nothing leaving the device for your CISO to review.

02

Your team clicks a workflow. Nobody writes a prompt.

More than half of employees have never written an AI prompt — and they won't. With 2,800+ pre-engineered workflows across sales, legal, HR, finance, healthcare, and government, every task is a button: pick the workflow, drop in the document, get expert output. That's why adoption hits 89% instead of stalling at 25% — an AI personal assistant for every employee, not a licence pool for the few who write prompts.

03

The output ships. The data never left.

Proposals, contract redlines, claim summaries, variance narratives — produced in seconds, on hardware you own, with zero bytes sent to any cloud. Add Blockify and the answers are grounded in your validated data with 78× fewer hallucinations. It compounds: every workflow your team runs is pure output, not a new line item.

The AirgapAI Philosophy
“What stays on the laptop, stays on the laptop. If your AI needs your data shipped to someone else's servers to be useful, you don't have an AI strategy — you have a data donation program.”
— The Iternal Technologies founding team
What's Inside

Everything in the box

Each component would cost more alone than the whole bundle does together. That's deliberate.

An institutional case lying open, its fitted tray cut with shaped recesses that are every one of them filled — a sealed module lit by a single cyan seam at the centre, refined blocks, a coiled conduit and further components, nothing left as an empty slot to buy later.

AirgapAI Chat Application

The full local AI assistant for Windows and macOS. Installs like any desktop app. Lifetime updates on perpetual licenses.

100% On-Device

2,800+ Done-For-You Workflows

Sales, marketing, legal, HR, finance, healthcare, insurance, government — engineered by domain experts, tested before shipping.

89% Adoption

Entourage Mode

@-mention @Legal, @Marketing, and @Finance in one chat and watch specialist personas produce one cross-functional answer — agent orchestration running entirely on the device.

3 Experts · 1 Chat

6+ Open-Source Local Models

Arcee 4.5B, LLAMA 3.2 (1B/3B), Gemma 9B, QWEN, and more via GGUF. Pick the right brain per task and hardware tier — small language models sized to run on an Intel NPU or integrated GPU.

Runs On iGPU/NPU

Local Document Ingestion

Drop in NDAs, contracts, transcripts, patient records — ask questions across them. Nothing is uploaded, ever.

0 Bytes Leave

Custom Workflow Builder

Describe a task in plain English and get a configured, reusable workflow back. Build unlimited internal automations.

Unlimited Builds

Compliance By Construction

SCIF-deployable, HIPAA / CMMC / NIST / FedRAMP / GDPR-residency aligned — because the data physically never leaves your control. Clinical teams can see how the same boundary produces HIPAA compliant AI.

CISO-Ready

Fleet Deployment & Governance

Push via Intune, JAMF, or SCCM. Role-based workflow libraries, approval gates, audit logs, version control.

10,000+ Seats

Blockify™ Add-On (Enterprise tier)

Your messy files become validated IdeaBlocks before the model sees them. Near-zero hallucinations on your own data.

78× Accuracy
Bought separately, this stack prices out at $15,000+
$697 once  ·  or $35/mo after free trial

AirgapAI vs. the cloud copilots

Two halves of one drawing: on the left a single solid block resting complete on its plinth, finished and fed by nothing; on the right a tap dripping steadily into a long receding row of separate cups, its supply line running away out of frame.
  AirgapAI Cloud AI subscriptions
Runs 100% on your device
Works offline — planes, SCIFs, remote sites
Your prompts stay out of third-party training pipelines
Price $697 onceor $35/mo, free 7-day trial $2,060+ per seat / 3 years (Forrester)
Unlimited prompts — no tokens, no overages
2,800+ pre-built workflows included
HIPAA / CMMC / FedRAMP posture by design requires contracts + review
Fleet deploy via Intune / JAMF / SCCM
Inside The Boundary

Secure AI chat for enterprise: what stays inside the boundary

Most security reviews come down to one question: which bytes cross the network. With AirgapAI the answer is a short, fixed list — and it is the same list whether you are running one laptop or ten thousand. If the term itself is new to your reviewers, start with what is air-gapped AI.

Stays on hardware you own
  • The model weights and every token of inference — generation happens on your CPU, GPU or NPU
  • Prompts, responses and chat history
  • Documents you ingest and the local index built from them
  • The 2,800+ shipped workflows and any workflow your team builds
  • Attachments a colleague would otherwise paste into a cloud chat window
Crosses the network only when you choose
  • The installer download, once per machine or once per image
  • Model packs you decide to add, on your schedule
  • Product updates, which you can stage and push like any other app
  • Nothing else: no telemetry, no prompt logging, no call-home check
  • Your content is never used to train a model — there is no pipeline that could carry it

Deploy and scope it like any managed app

Push AirgapAI through Microsoft Intune, JAMF or SCCM, then scope workflow libraries by role so finance, legal and engineering each see their own set. No new identity system to stand up, and teams that would rather hand the rollout and update cycle to Iternal can run it as managed AI services.

Governance that survives an audit

Approval gates on published workflows, version control on the ones you build, and usage reporting by department — the evidence CMMC, HIPAA and FedRAMP reviewers ask for, without a third-party processor in the diagram.

Answers grounded in approved content

With the Blockify add-on, source files become validated IdeaBlocks before the model can read them, so a secure chat is also an accurate one — a documented 78× aggregate accuracy gain on enterprise data.

The alternative is the ungoverned path: IBM's 2025 Cost of a Data Breach report puts breaches involving ungoverned AI tools at roughly $4.6M on average (IBM, 2025). For the wider category, see private AI and private LLM deployment; for a build-it-yourself conversational rollout, see AI chatbot development services. For a ranked look at the wider field, see best AI for air-gapped environments, and for the job-by-job version, see Running AI on Data That Cannot Go to the Cloud.

Get the Security & Deployment Guide Or score your own position with the AI data privacy assessment and the secure AI ROI calculator.
Field Results

$650M in pipeline. Proposals cut from 6 weeks to 2 days.

Fortune 500 sales teams use AirgapAI to draft proposals, research accounts, and rip through RFPs at 10× speed — generating $650M+ in net-new pipeline while competitors copy-paste deal data into ChatGPT and pray nothing leaks. That copy-paste habit is the textbook shadow AI risk: sensitive data quietly leaving the network with no audit trail.

$650M+
Net-new pipeline
10×
More shots on goal
6w→2d
Proposal cycle time
Our
Better-Than-Free
Guarantee

Try it free for 7 days. The risk lives with us, not you.

You get full access for a week. No charge, no credit card hold, no "starter tier." If by day 7 it isn't measurably saving your team time, cancel inside the app and we never charge you a cent.

Buy the perpetual license? Same deal. Use it for 30 days — if it doesn't pay for itself, email us and we refund every penny. We'd rather lose a sale than have a customer regret one. Fewer than 0.5% of buyers ever ask.

  • 7-day free trial — $0 today, full feature access
  • 30-day money-back on perpetual licenses
  • Cancel inside the app — no calls, no friction
  • No data hostage situation — it's all on your machine anyway

Not ready to install anything? Two zero-commitment next steps.

Get the architecture in front of your CISO, or tell us about your environment and we'll tailor a demo to it.

Take it to your CISO first

The Security & Deployment Guide: data-flow diagram, SCIF/CMMC/HIPAA/FedRAMP compliance mapping, and a 10-point evaluation checklist.

Get the Security Guide

Evaluating for a fleet, SCIF, or agency?

Tell us about your environment — we'll tailor a 15-minute demo with the right models, workflows, and deployment path for your rollout.

Book a Tailored Demo
Frequently Asked Questions

Questions buyers ask before pulling the trigger

If your question isn't here, ask in the site chat for an instant answer or email support@iternal.ai.

What is AirgapAI?
AirgapAI is a 100% local AI assistant from Iternal Technologies. The models, 2,800+ pre-built workflows, and your documents all run and stay on your own Windows or macOS laptop — no cloud services, no internet connection, and zero data transmission. It is priced at $35/month after a 7-day free trial, or $697 one-time for a perpetual license.
How is this actually different from ChatGPT or Copilot?
ChatGPT and Copilot run in someone else's cloud. Every prompt you send leaves your network. AirgapAI runs entirely on your laptop — the model, the workflows, the data. No internet required. That's why it's approved for SCIFs and classified work, and why $697 once beats $2,060 over 3 years (Forrester). See the full edge AI vs cloud AI cost breakdown, or the enterprise AI strategy guide for where local AI fits in a broader rollout.
Is $697 really cheaper than ChatGPT Enterprise long-term?
By a lot. ChatGPT Enterprise and Copilot run $30+/user/month — Forrester puts the loaded 3-year cost at $2,060+ per seat. AirgapAI is $697 once, with lifetime updates. Across a 1,000-person company, you save more than $1.36M in 3 years and you get to deploy AI to everyone, not just the 20% you could afford before.
Will my employees actually use this?
Most enterprise AI tools die at 20–30% adoption because users don't know how to prompt. AirgapAI hits ~89% adoption because every workflow is a button. Your accountant doesn't need to learn prompt engineering — they click "Variance Narrative," paste their P&L, and ship the output. Day one.
What about hallucinations? Can I trust the answers?
Out of the box, AirgapAI runs the same class of open models as the major clouds. Add the Blockify upgrade ($1,499 Enterprise tier) and independent benchmarks show 78× fewer hallucinations on your own data — because Blockify pre-validates every chunk before the model can touch it.
Is it really approved for classified / SCIF environments?
Yes. Because AirgapAI never calls home, it satisfies the air-gap requirement by construction. It's deployed inside SCIFs, on classified Dell + Intel laptops, and in DoD-adjacent workloads today. HIPAA, CMMC/NIST, FedRAMP, and GDPR data residency are all compliant by design — see the AI compliance frameworks overview, plus EU AI Act Article 4 requirements for AI literacy.
Does AirgapAI work offline with no internet connection?
Yes. AirgapAI performs all inference, document ingestion, and workflow execution on-device, so it works on planes, in SCIFs, at remote sites, and anywhere else without connectivity. You can unplug the network cable mid-answer and it keeps generating — nothing requires a connection except downloading the installer and updates.
Is AirgapAI a private assistant, or does anything sync to a server?
Nothing syncs. AirgapAI is a private AI assistant in the literal sense: the model weights, your prompts, your ingested documents, the local index and your chat history all sit on the machine you own. There is no account sync, no telemetry, and no call-home check — the only traffic is the installer, model packs you choose to add, and updates you decide to install. For the category background, see what private AI is and private LLM deployment options.
What makes it secure AI chat for an enterprise fleet, not just one laptop?
The same boundary applies at 10,000 seats as at one. You push AirgapAI through Intune, JAMF or SCCM, scope workflow libraries by role, gate which workflows get published, version-control the ones your team builds, and report usage by department — with no third-party processor in the data-flow diagram, which is what shortens CMMC, HIPAA and FedRAMP reviews. The Security & Deployment Guide has the diagram and the 10-point evaluation checklist, and the boundary section above lists what does and does not cross the network.
What hardware do I actually need?
Windows 10/11 with an Intel Core Ultra CPU or any dedicated NVIDIA/AMD GPU works. macOS needs an Apple M1 or newer with 8GB+ memory. The bigger your GPU/NPU, the bigger the model you can run — see the full system requirements below for generation-speed benchmarks.
Which AI models are included?
6+ open-source models ship in the box: Arcee 4.5B, LLAMA 3.2 (1B and 3B), Gemma 9B, QWEN, and more — with additional open-source models supported via the GGUF format. You pick the right model per task and hardware tier, and switch any time. No per-model fees.
Can I deploy it to my whole company?
Yes. Push it through Microsoft Intune, JAMF, or SCCM — the same MDM tooling you use for everything else. Lock down workflows by role, audit usage by department, version-control your custom workflows. We routinely roll AirgapAI out to 10,000+ seats. For volume licensing, POs, or a tailored rollout plan, start with the 2-minute discovery questionnaire.
What if it doesn't work for us?
You cancel during the 7-day free trial and pay nothing. Or you buy the perpetual and request a refund inside 30 days. We'd rather you walk away clean than feel stuck. The risk is on us — see the guarantee above.
More From AirgapAI

The rest of the local-AI toolkit

AirgapAI Transcribe

100% local AI transcription + meeting notes. Audio never leaves the device. $29/mo (7-day trial) or $599 once.

Explore Transcribe

AirgapAI Code

Secure agentic coding that never leaks your source. Runs inside your boundary. From $80/mo (7-day trial).

Explore Code

Stop renting your AI. Start owning it.

Every month you wait is another $30/seat handed to a cloud provider that's training its next model on your data. Try AirgapAI free for 7 days. If we don't earn it, you don't pay it.

Start My 7-Day Free Trial — $0 Today
SCIF-approved 100% local Cancel anytime 30-day refund on perpetual

AirgapAI runs large language models 100% locally on your hardware. Available for both Windows and macOS, performance varies based on your device's GPU capabilities — better hardware means faster responses.

Windows: Running LLAMA 3.2 1B on Intel Core Ultra 7 165U with integrated graphics achieves approximately 100+ words per minute generation speed. An Intel 258V generates 1,800 words per minute.

macOS: Apple Silicon (M1/M2/M3/M4) provides excellent performance with unified memory architecture. M1 with 8GB+ RAM is the minimum requirement with 900+ words per minute generation speed.

MinimumGoodBetterBestUltra
Intel 165U/HIntel 165HIntel 258VIntel 265VIntel 268V
iGPU/NPUiGPU/NPUiGPU/NPUiGPU/NPUiGPU/NPU
16GB RAM32GB RAM64GB RAM64GB RAM64GB RAM
LLAMA 1BLLAMA 1B/3BLLAMA 3B/8B+LLAMA 3B/8B+LLAMA 3B/8B+

Windows Requirements

  • Large Language Model Inference Requirements:
    • LLAMA 3.2 1B — 2024 Intel HD Integrated Graphics, 8GB vRAM minimum
    • LLAMA 3.2 3B — 2025 Intel HD Integrated Graphics, 8GB vRAM minimum
    • Arcee 4.5B (2025) — Intel Core Ultra or dedicated GPU, 8GB vRAM minimum
    • GEMMA 9B — NVIDIA RTX 2000 or AMD equivalent
    • QWEN 2026 — NVIDIA RTX 2000 or AMD equivalent
    • LLAMA 3 8B — NVIDIA RTX 500 or AMD equivalent
    • Additional open-source models supported via GGUF format
  • Supported Operating Systems: Windows 10 and Windows 11

macOS Requirements

  • Minimum: Apple M1 with 8GB+ of Memory
  • Recommended: Apple M1 Pro/Max/Ultra or M2/M3/M4 series with 16GB+ of Memory for larger models
  • Supported Operating Systems: macOS 12 Monterey or later