01 · Quotes
Describe the job. Send the quote.
A quote on your letterhead with GST and your rules checked, produced in the conversation and saved as a PDF ready to send.
On-premise AI for Australian business
Your AI. Your machine.
Your office.
Search everything your company already knows, turn meetings into action, research a question properly and hand the routine work over — private, on-premise, on hardware you own outright. We build the software, fit the machine and hand it over working.
We are ready to start a closed pilot of the platform with businesses. On-site installs and demonstrations are coming soon; register your interest now.
Register interest for a pilot See what it can do- No per-user subscription
- Base local work stays on your network
- Installed and handed over
On-site installation in our listed NSW regions; configured supply and support Australia-wide.
Ask once.
Get the work back.
The Harness runs on hardware you own. Connect your sources, choose the models and decide who can see what. We install it, hand it over and stay nearby.
02 · Invoices
Turn invoices into checked rows.
Where invoice capture is included in the written scope, we fit the extraction workflow to your documents and demonstrate it with agreed samples before handover.
03 · Meetings
Meetings that end in next steps, not another set of notes
Record a meeting on the office network. Get the summary, the decisions and the actions with owners and due dates.
The five to test first
What your team will actually use.
These common office jobs are straightforward to test against agreed examples before you buy. There are more — the full feature list runs to fifteen sections.
$0
Harness fee per user, per month
We do not charge a Harness fee for each staff account. Machine capacity still limits how much work can run at once, so the quote sizes the hardware for expected use.
Base offline work stays on your appliance
Prompts, documents, recordings and the model work all stay on the appliance. Anything that would leave is a switch you turn on yourself, named in writing.
Answers with the working shown
Ask about a spreadsheet on your own file store. You get the answer, the steps it took, the source file — and a chart if you want one.
Meetings that end in next steps
The transcript becomes the summary, the task list, the patient file — whatever your office writes up after a conversation.
Every meeting, searchable
“What did we agree with the builder in March?” becomes a question with a cited answer, not an afternoon in the archive.
One question becomes seven
Deep research splits the question into separate lines of enquiry, reads across your library and the web where you allow it, and writes a report with every figure cited.
Standing jobs, when included
A morning brief, a Monday digest, an aged-debt list. Scheduled work is listed in the quote and must pass its delivery-hardware acceptance test before handover.
Project records become a handover pack
Bring drawings, meeting notes and defect records together. The Harness separates open items, cites the supporting record and writes the pack for review.
Own the working system
Why businesses choose LocallyAI over ChatGPT or Claude.
Hosted assistants process work inside provider infrastructure your business does not own or operate. LocallyAI gives approved routine work a defined path on hardware you own, with storage, accounts, models and connected exceptions documented at handover.
Processing location is the difference
Hosted services require your business to rely on a provider’s infrastructure, terms, configuration and continuing service. LocallyAI keeps approved base workflows on your appliance; every connected exception is named in writing.
A boundary your team can inspect
The standard workflow is local. Web research, integrations, remote support, updates and off-site backups are separate exceptions recorded in the quote and tested before handover.
Core work can continue offline
Local chat, approved document search and transcription can keep running when the connection drops. Web research and connected services cannot.
Built from scratch, for this
We did not rebrand a coding interface. The Harness is written for office workflows and fitted to the templates, letterhead and approved systems named in your quote.
Connections are quoted, not assumed
Where a supported connection is needed, its scope, permissions and test cases are written into the quote. No integration is treated as delivered until it passes on the final machine.
You can actually talk to us
We build the software, so a requested change can be discussed with the people who write it. Standard product improvements and business-specific work are separated in the written scope, including any cost.
Quoted automations run on the appliance
A morning brief, weekly digest or aged-debt list is included only when named in the quote and proven against agreed examples before handover.
The Harness is modular
Supported skills and MCP connections can be added without replacing the interface. Each requested module still needs a written scope and delivery test.
See everything included in the LocallyAI Harness →
Why a Harness at all
A model is a horse. The Harness is the saddle.
The models everyone talks about are horses — powerful, and useless to a business on their own. The saddle is what makes one rideable: the software that straps it to your documents, your templates, your rules and your people. Nearly every saddle ever built was made for programmers. The LocallyAI Harness is built for running a business.
One Harness. Hardware sized for the work.
Three good ways to run it.
The Harness and the machine are one system — never a bare licence, never a bare box. Software, setup and support are the same on all three paths; the only choice is which machine. We recommend a Mac for most offices because it is quiet and simple to live with. Strix Halo and DGX Spark are first-class when Windows, Linux or CUDA matters more.
Recommended for most offices
Mac
Mac mini and Mac Studio systems are quiet, efficient and straightforward to keep serving models in the background. Setup is uncomplicated and support is easiest.
Windows or Linux
Strix Halo
Compact AMD machines with up to 128 GB of unified memory, strong local-model performance and a familiar PC environment for offices that need one.
NVIDIA and CUDA
DGX Spark
A 128 GB Linux system for CUDA software, AI development and larger-model work. Two can be paired for selected workloads.
We size the machine after seeing the work, not from a showroom. Compare the three hardware paths →
The privacy boundary
Local by default, with the exceptions named.
In the standard offline configuration, prompts, documents, recordings and model work stay on the customer appliance. No cloud account, no per-seat fee and no data sent away for the system to function.
Data leaves only through a feature you separately enable or request: web research, an integration, an update, remote support or off-site fine-tuning. Your quote and setup record name the exceptions you have chosen, in writing.
A buyer’s shortcut
When owning the machine makes sense.
Good fit
Several people repeat work involving client files, long documents, recordings or private internal knowledge. You want predictable ownership, no per-person licence and a written processing boundary.
A hosted service may be simpler
Use is occasional, the material is public or low-risk, or your priority is immediate access to the newest frontier model. We will say so when a machine would be more system than the job needs.
Installed pricing
Clear starting points. A written quote for the real job.
The price includes the selected computer, the Harness, local models, setup and handover. We confirm the exact platform after seeing what it needs to do. Ten seats at $30 a month is $18,000 over five years, and you own nothing at the end of it — the machine is bought once.
- Micro · one person, dictation and drafting
- from $2,900
- Compact · one or two regular users
- from $4,900
- Office · a small shared team
- from $7,900
- Practice · heavier models or more people
- from $10,900
- Firm · high memory or paired systems
- from $17,500
Prices are in Australian dollars and exclude GST. LocallyAI is registered for GST; how quotes show it is explained on the pricing page. Hardware pricing moves, so written quotes are valid for 14 days.
Sized to the work, not to the ceiling.
These are planning tiers, not automatic recommendations. If a 16 GB Mac running a small model and transcription passes the agreed workload, we will quote that instead. The right machine is the one that matches the work — not the largest one we could sell you.
The Harness
Built by us, improved continuously.
We wrote the Harness from scratch — it is not a reskinned coding tool — so improvements reach your machine through the support plan rather than a licence renegotiation. Applying an update needs a temporary connection; the normal local setup goes back offline once it is done.
A sensible first step
Show us one annoying job.
Bring a redacted document, a non-sensitive recording or a task your team repeats every week. We will show you how the Harness handles it and tell you plainly which machine makes sense — including when the answer is that you do not need one.
We are ready to start a closed pilot of the platform with businesses. On-site installs and demonstrations are coming soon; register your interest now.
Register interest for a pilot How an install works
See what LocallyAI could take off your team’s plate.
Private. Fitted to real work. Running on a machine in your building.
We are ready to start a closed pilot of the platform with businesses. On-site installs and demonstrations are coming soon; register your interest now.
Register interest for a pilotQuestions people ask first
The short version.
Is this like ChatGPT?
It feels familiar: open the Harness in a browser and ask for help. The difference is where the work runs. In the standard setup, the model and your office files live on a machine you own.
What is the simplest AI setup for a small business?
One machine on the office network, reached in a browser. The Micro — a 16 GB Mac mini from $2,900 — can suit one person using dictation, approved document search and drafting. The written scope records the final machine, network prerequisites and installation window.
Can I buy the Harness for a computer I already own?
Yes — bring us the machine and we will assess what it can genuinely run. If it suits the work, the hardware line comes off the quote and you pay for the software and configuration only. If it is not up to the job, we will say so plainly. The Harness is never sold as a bare licence: whichever machine it lands on, we configure, test and hand over a working system.
Does it need the internet?
The core assistant, document search and transcription can run without it. Web research, integrations, updates and remote support need a connection when you choose to use them.
Can we choose Mac, Windows or Linux?
Yes. We usually recommend a Mac because setup and day-to-day model serving are straightforward, but we also fit AMD Strix Halo machines for Windows or Linux and NVIDIA DGX Spark systems for CUDA and larger-model work.
How do Harness updates work?
The Harness is ours, so improvements reach your machine without a licence renegotiation. Applying an update needs a temporary connection; the machine returns to its normal offline setup afterwards.
What does an installed system cost?
Installed systems start at $2,900. The final price depends on memory, storage, operating system, the models you want to run and the work we configure. Prices are in Australian dollars and exclude GST. LocallyAI is registered for GST; the pricing page explains how quotes show it.
Can we see it before deciding?
Yes. We can walk through the real Harness with example documents and show the jobs that matter to your team. A first chat is free and there is no obligation.
Do you supply and support businesses anywhere in Australia?
Yes. We quote on-site installation in our listed New South Wales service regions. Elsewhere in Australia, the normal route is a configured and tested appliance shipped to your office, followed by setup with you and your IT provider over a temporary remote connection. On-site travel can also be discussed where practical.
Can we supply our own machine?
Yes. Bring us the machine and we will assess what it can genuinely run — memory, storage, operating system, and how much headroom is left once a model is loaded.
If it suits the work, the hardware line comes off the quote entirely and you pay for the software and configuration only: the Harness, the local models, your own templates and the setup. If it is not up to the job, we will say so plainly and tell you what it would take, rather than install something that will disappoint you.