ONYRIX
Get a use case Request a pilot
Limited pilot program · First year free

Patent pending

Prove the work.
Own the record.

An AI-powered workspace where your team delivers around the clock — and every task is verified the moment it’s done, whether a person or an AI agent did it. No status meetings. No self-reported progress. Proof that travels with the people who earned it.

Watch the overview Request a pilot
MIC Market OC Organisation PDC Devices Virtual workspaces SFO NYC LDN BER TOK SYD
Always on — the 24-hour company architecture.

The problem

Work history runs on trust,
not proof.

Résumés are claims. References are slow and easy to game. Companies can’t truly verify who did what — and the people who did the work can’t carry that proof anywhere they go. Everyone’s relying on trust where there should be evidence.

And now some of the work isn’t human. AI agents are writing your status updates, your handoff notes, and making changes directly in your systems. They produce a confident sentence whether it rests on something that actually happened or on nothing at all — and the output doesn’t tell you which. Same problem, new shape: a record everybody has to trust because nobody can check it.

What you get

Delivery you can watch.
Work you can prove.

Onyrix replaces the reporting layer with evidence. Three things change from the first week of a pilot.

01

The work doesn’t stop overnight

A task picked up in London continues in New York and finishes in Sydney. The system writes the handoff at the shift boundary itself — so nobody burns a morning working out where the last person left off.

02

You stop asking what got done

Every completed task is verified as it completes and written to your company ledger. Progress you can look at, instead of progress somebody told you about in a meeting.

03

You staff on proof, not résumés

Match to professionals on verified work in the exact category you need, attested by the companies they did it for — not on a claim you’d have to take on faith and a reference call you can’t validate.

Who it’s for

Built first for the businesses
that get paid on other people’s work.

If your revenue depends on work performed by people you don’t watch — contractors, placements, distributed teams, and now AI agents — you are carrying a risk you currently have no way to measure.

01

Payroll & workforce companies

You process what gets reported to you. You can’t confirm any of it. Disputed hours, ghost time and audit exposure all trace back to the same gap — the hours arrive as a claim, not as evidence.

Onyrix makes the hours arrive with proof attached.

02

Staffing & recruiting firms

Your margin rests on placing people who can actually do the work. Today that judgment runs on a résumé you can’t check and reference calls that tell you nothing. A bad placement costs you the fee, the client and the rebate.

Place on verified history. Prove delivery to the client.

03

Software & technical services

Distributed teams, work crossing time zones nightly, and AI agents already touching your systems. Nobody can tell you what an agent actually changed, or whether the handoff note it wrote was based on anything real.

Continuity that carries context, and agents you can audit.

The overview

See it before you commit
twenty minutes to a call.

Two minutes, end to end: how work gets verified as it’s done, matched to the right people, and carried across time zones without ever stopping.

Watch the 2-min overview

The workspace

An AI-powered virtual workspace where the work actually happens.

Your project runs in a shared, AI-assisted room. An AI intermediary keeps work moving — flagging duplicated effort, surfacing decisions that are blocking progress, and carrying context across every shift. Nothing is self-reported: every completed task is verified into your company ledger.

London→ New York→ Sydney

Work that never sleeps.

A task picked up in one region is handed off to the next — the workspace carries the context, so delivery continues around the clock.

  • Verified delivery. Every completed task is proven, not self-reported.
  • 24-hour continuity. Progress doesn't stop when a region logs off.
  • AI keeps it moving. Duplication flagged, blockers surfaced, context preserved.

Representative illustration of the virtual workspace. Patent pending.

Evidence-grounded records

Every line tells you
where it came from.

A handoff note here isn’t a paragraph of AI-written prose. It’s a list of individual statements, and every one of them is labelled — in the record, in an export, and when you ask about it in conversation. An inference can never be dressed up as an observation.

Observed
It’s in the workspace log. The system can summarise what happened. It cannot go one inch past it.
Attributed
A person said it, and it’s shown as their words — never as the system’s finding. That someone said the blocker was cleared is a fact. That it was cleared is a different fact, and the system won’t assert it for them.
Inferred
The system worked it out — and it carries the observations it worked from, so you can trace any conclusion back to the ground it stands on.
Judgment
A recommendation with nothing behind it yet. Allowed in “what I’d do next.” Barred from every field that describes the state of the work.

If something can’t be backed up, it’s marked as an inference with its reasoning attached, or it’s dropped — and the drop is recorded too, pointing at what was left unresolved. A note with a visible gap in it beats one that reads beautifully and is wrong.

Then it gets checked from the other end. The person handing off reviews the record but can’t rewrite it — they raise an objection and the system re-checks its own work. The person picking it up marks each statement against what they actually found on arrival. What builds up is a running account of what each person was told and what the next one found to be true — the documentation you go looking for when a job goes sideways and nobody can remember who knew what.

AI agents at work

Put AI agents on real work — without handing over the keys.

Everyone is being asked to let AI into the systems that run the business, and nobody wants to be the person who signed off on it. Onyrix lets you say yes. An agent joins the workspace as a credentialed participant with its own verified identity — the same boundaries a person gets, enforced the same way.

It can’t widen its own access

You set the ceiling once: which kinds of records, which kinds of operations, read or write, reversible or not. Anything outside it is refused before it ever reaches the agent. Not a rule the model was asked to respect — a door that isn’t there.

You can pull the plug mid-task

Cut access for one agent, a team, or the whole workspace, and it lands inside sessions that are already running. Anything caught in flight is brought to a clean stopping point and recorded — never left half-finished with nobody sure what happened.

The audit trail is written first

What an agent is about to do is committed to the record before it acts, not reported back afterwards. If it fails, stalls, or gets cut off mid-operation, you still have the evidence of what was attempted.

Even the software with no integration

Onyrix learns what an application can actually do — including the old system nobody wants to touch, whose vendor stopped developing it years ago — and puts the same governed, audited boundary around it. No vendor involvement required.

And it’s graded on results, not on how confident it sounded.

Work an AI produces here is scored against what independently happened next — did the change actually take effect, did the next person find it accurate, did the client sign off. Scored separately for each kind of work, because something dependable at one job isn’t automatically dependable at another. When a capability starts getting things wrong, the system does less of it, flags what it does produce, and can withdraw it entirely — automatically, everywhere, at once.

Governed by the Virtual API Security Layer — scoped access, instant revocation, audited before the fact.

What it is

An AI-powered distributed work platform,
built on verification.

Every task a professional completes is recorded on their own device, summarised upward to the companies they work for, and — only in anonymised form — to a market-wide intelligence layer. Professionals own their history. Companies get proof. Nothing crosses a tier it wasn't meant to.

The architecture

Three tiers. One direction of trust.

Proof is built at the edge and rolls up as cryptographic summaries — never raw data. Privacy isn't a layer on top. It's the direction the data moves.

PDC

Personal Device Chain

Your verified résumé, built as you work. Records are generated on your own device, owned by you, and carried from job to job.

OC

Organisational Chain

Each company's ledger of verified, completed work — built from cryptographic summaries rolled up from professionals' devices.

MIC

Market Intelligence Chain

Know where you stand in your market. Anonymised, market-wide signal on skills and demand — no personal data, no company secrets.

How it works

The 24-hour company architecture

VERIFY. MATCH. DELIVER

01

Verify

Work is proven cryptographically as it's completed, on the professional's device — not self-reported after the fact.

02

Match

An AI engine matches verified professionals to the companies that need them — on proven capability, not just what's claimed on paper.

03

Deliver

Work happens in shared virtual workspaces with AI-assisted handoffs, so a task picked up in one time zone continues in the next.

Always-on by design. Continuity isn't a setting you switch on — the verified, AI-powered handoff record is how the system works, so delivery doesn't stop when one region logs off.

Before you ask

The five questions everyone asks.

Do we have to rip out the tools we already use?

No. Onyrix is where the work is coordinated and verified — it governs access to the systems you already run rather than replacing them, including the ones with no integration path. A pilot starts around one team and one project, not a migration.

Who owns the work records?

The professional owns their own record, generated on their own device, and carries it from job to job. Your company ledger holds verified summaries of the work done for you. Neither one is the platform’s to sell out from under you.

What leaves our company?

Only anonymised, aggregate signal reaches the market layer — the shape of skills and demand across the market, never who did what for whom. No personal data, no company detail, no client names. It is a one-way direction of travel and it is built into the architecture rather than configured on top of it.

What does it cost?

Founding pilots run free for a full year. Pricing after that is set together with the founding cohort, once a year of real usage has shown what the thing is actually worth — not guessed at beforehand.

How much of our time does a pilot take?

A 20-minute call to find out where verifying work is costing you most. Then we build the workspace around one real team and one real project. You run it with your people; we build alongside you.

No call required

Not ready to book a meeting?
Get the use case in writing.

Plenty of people want to know whether something is worth their time before they give anyone twenty minutes of it. Fair. So here is the other way in.

Tell us how your operation actually runs.

Where work gets handed off. Where it goes missing. What you currently have no way to prove. Six questions plus your email — write as much or as little as you like.

  • We write back with a use case built on your answers — what Onyrix would do in your business, with the parts that don’t apply left out
  • If there’s no fit, we’ll tell you that instead — and say why
Get a written use case No calls — just email.

Who’s building this

You’d be talking to me.

Phillip Meyer, founder of Onyrix Inc.

I spent about seven years estimating commercial HVAC in Atlanta — somewhere near 5,700 bids, including the one for the Porsche North America headquarters. Before that I grew up in my family’s industrial automation business in South Africa: robotics, welding and safety systems for car plants, until I moved to the US.

Two things stuck. Estimating teaches you that the difference between a job that makes money and one that doesn’t is almost always something nobody wrote down. And every time I changed companies, everything I’d proven about myself stayed behind. The work was real. The record of it was never mine, and the next employer had no way to check a word of it.

Onyrix started as that question — why can’t a person own proof of their own work? Then AI agents started doing the work too, and the question got a good deal bigger.

I’m building this full time, seven days a week. If you run a pilot, you’re dealing with me directly.

Phillip Meyer Founder, Onyrix Inc. · Atlanta, GA

Limited pilot program

Join a small group of founding companies.

We're accepting a limited number of companies into the first Onyrix pilot cohort. Founding pilots help shape the product — and get in before anyone else.

Your first year is free — founding pilots only.
01

A 20-minute call

Tell us where verifying work costs your team the most. No prep, no commitment.

02

A tailored pilot

We build a workspace around one real team and one real project.

03

A full year, free

Run it with your team at no cost while we build alongside you.

And if it isn’t working at 90 days, we shut it down and you keep every record it produced. No notice period, no exit conversation, nothing to unwind.

Request a pilot

Not ready for a call? Get a written use case instead — no meeting.

Limited spots — the founding cohort closes once it’s full. Pricing after the pilot year is set with the cohort, not before it.