inference · for companies
inference · for companies
NFRNS, said “inference” · on-premises edition coming soon
NFRNS reads how your business actually works, from your own documents and systems, maps the processes, and proposes what to automate. Then it runs them, stopping for a yes at every step that matters. Nothing leaves your building. Nothing runs without a person.
One Linux box with Docker. Open model, Apache-2.0. Your data never leaves it.
The AI assistant you message on WhatsApp that does the job: finds it, books it, pays for it, remembers what you told it. Running on NFRNS’s own servers in the UAE; on your device soon. heysam.bot
Learns your business from your own documents and systems, proposes what to automate and runs it with your approval. On your premises, or hosted by NFRNS. This page.
A folder of handbooks and policies, a shared drive, exports from your systems. PDF, Word, spreadsheets, email. Read and indexed on your own machine.
Who does what, when, in which system, with which handoffs and pain points. Where the documents are silent, it asks the owner.
What an assistant could take over, with a risk class and an approval gate on every step that sends, pays, orders or changes a system.
Nothing runs before someone says so. Approvals, edits and rejections go on an audit chain you can verify offline.
Playbooks run on your schedule or on demand, in chat or in the console, and stop for your yes at the gates. Every run is on the chain.
Every playbook NFRNS proposes comes with the evidence it was built from and a gate on each step that sends, pays, orders or changes a system. A person approves, edits or rejects it in the console or in chat. Every decision, and every run, goes on a hash chain you can verify offline, long after the people who made it have moved on.
NFRNS console · Acme Trading · on-premisesFound in: Finance handbook, Expense policy 2026. Claims are filed late and receipts go missing.
Your people want an AI that knows them. You want an AI that knows the company. NFRNS gives each its own brain and puts a membrane between them, so neither leaks into the other.
Your documents, your processes, your playbooks, on your hardware. Readable by role, so the board pack stays with the board.
Every exchange between a person’s brain and the company’s is parsed, labelled personal, work or sensitive, redacted, and allowed or refused by the grants you set. Offboarding is one step: the link is revoked and company data is cleared from the person’s brain.
SAM, our assistant for people, kept separate: the same person, their own brain, their own life. Cards, IDs, codes and credentials never cross, whatever the grants say.
One Linux box with Docker: the model, the brain and the console. 64 GB of RAM runs the open model on CPU; a 24 GB GPU makes it fast. ./nfrns init, ./nfrns up, ./nfrns connect, ./nfrns discover. No outbound traffic once the model is on disk.
The same brain on NFRNS’s own servers in the UAE, one tenant per company, each sealed with its own key. For teams that want it running this month, before the on-premises edition ships.
An assistant that knows your product, your customers and your runbooks, without wiring up a cloud stack: the same box, your data, a console your team uses on day one, and an API your product calls. Open model, open weights, no per-token bill to a model vendor.
Tell us what you would automate first. We onboard a few companies at a time and set the first one up with you.