Payroll intelligence and audit, built on the data already in your ERP.
Press P to project · Arrows or space to advance · Esc to exitIt isn't carelessness: checking 1,200 employment records rule by rule, every month, doesn't fit inside a human team. So payroll gets spot-checked — and the error only surfaces once it's already a liability, with a legal clock running against the company.
All three deadlines are public and verifiable. What's missing isn't the law — it's somebody checking before the clock runs out.
That doctrine governs every line of the product. No number here comes out of a language model: the math is deterministic and reproducible. The AI writes the prose that explains the number. And nothing gets published, sent, or executed until someone says yes.
145 payroll checks and 40 timekeeping checks, all deterministic. Run it twice, get the same answer.
The book's prose, the reading of a metric, the explanation of a finding. Never the value.
New queries, book delivery, source publication. Every one goes through a person.
The product menu, the Book's notebooks and this presentation use the same names.
Every seat at the table has its subject — <b>and a notebook with its name on it.</b>
| 78 |
the score starts at <b>100</b> and comes down factor by factor, each one's cap written beside it
The score arrives <b>with the ruler attached.</b>
Labour code art. 477 §6 · one salary penalty · with the file line where it sits
The error surfaces on Wednesday. <b>Payday goes out right on Friday.</b>
| Base pay | R$ 17.00 |
|---|---|
| Payroll taxes | R$ 6.26 |
| Benefits | R$ 3.70 |
| Accruals | R$ 1.43 |
R$ 17.00 × 1.67 = R$ 28.39 per hour · × 220 hours × 1,248 people = R$ 7.80M — difference R$ 0.00
Every number ties to the one above it — <b>and that reconciliation runs as a test every day.</b>
The charts below are drawn by the same functions that draw the customer's report and book. This isn't a picture of the product — it is the product.
| Sep 25 | R$ 6.67M |
|---|---|
| Oct 25 | R$ 6.80M |
| Nov 25 | R$ 6.74M |
| Dec 25 | R$ 6.99M |
| Jan 26 | R$ 7.37M |
| Feb 26 | R$ 9.88M |
| Mar 26 | R$ 7.02M |
| Apr 26 | R$ 7.12M |
| May 26 | R$ 7.08M |
| Jun 26 | R$ 7.47M |
| Jul 26 | R$ 7.67M |
| Aug 26 | R$ 7.80M |
| HQ São Paulo | 486 |
|---|---|
| Site 02 Rio | 322 |
| Site 03 Minas | 244 |
| Site 04 Paraná | 142 |
| Site 05 Bahia | 54 |
| no accepted eSocial filing | 3 |
|---|---|
| Distribution of active headcount. | 1,248 |
| Operations | Mon: 34, Tue: 18, Wed: 9, Thu: 11, Fri: 22, Sat: 6 |
|---|---|
| Sales | Mon: 12, Tue: 7, Wed: 5, Thu: 4, Fri: 15, Sat: 2 |
| Maintenance | Mon: 9, Tue: 6, Wed: 8, Thu: 5, Fri: 10, Sat: 4 |
| Administration | Mon: 4, Tue: 3, Wed: 2, Thu: 3, Fri: 5, Sat: 0 |
every open item ships with the government's own guidance beside it, and a plan with a deadline and an owner
The liability shows up in currency, with the legal source beside it — <b>and a plan with a deadline.</b>
The most tangible thing the product makes: the Timekeeping Compliance Report and the Delivery Book. Both ship in a light theme, ready to print and hand to the board.
A signal is not a verdict: each finding is a lead to investigate, with its legal basis and evidence.
The documents shown are facsimiles built from fictional data.
scoreboard of the last 8 quarters: 7 landed inside the band — and the one that did not is written down
Forecast and actual, side by side: <b>this vendor plays with the scoreboard on.</b>
| 55+ | 38 · 34 |
|---|---|
| 45-54 | 96 · 84 |
| 35-44 | 178 · 163 |
| 25-34 | 214 · 206 |
| 18-24 | 128 · 107 |
Demographics stop being an appendix — <b>they become the decision about who you need to keep.</b>
every marker with its auditable origin in your ERP — in an inspection, this is evidence
The salary line tells the career — <b>with the origin of every fact.</b>
| site | employees | expires in |
|---|---|---|
| 01 · Head office | 8 | 18 days |
| 02 · Site | 4 | 25 days |
| 03 · Site | 2 | 33 days |
Every new query waits for a person to approve — <b>the pause is the answer.</b>
Eleven fronts sharing the same data, the same dictionary, and the same audit trail.
rescisoes.json · line 218147 deterministic checks on every close — arithmetic, employer taxes, terminations, job protections, vacation, 13th salary, and the payroll-code bindings the ERP close requires.
Ask in plain language. Opyra resolves the query, routes it for human approval when it's new, and returns the number with the full trail attached.
41 checks across the four official Brazilian clock-in file formats. A score per site, findings with legal basis, and a 30/60/90 remediation plan.
A conformity index over your real filing picture in eSocial (Brazil's unified payroll reporting system): what went through, what bounced, what nobody ever sent.
One person's entire working life on a timeline — hire, transfers, promotions, raises, leave, and exit, each fact showing where it came from.
An opening, four subject notebooks and appendices, every pay period. The rule calculates, the AI writes, a human approves before it ships.
Every answer Opyra knows how to give, browsable by the management question behind it — and which ERP field switches each one on.
Accident-risk and social-fund rates checked against the official tables. The gap comes out in reais.
Ask the union agreement, the contract, the amendment. It answers only from your documents — and cites the page.
The query you run every 5th of the month starts arriving on its own. Set a target on any metric and get an alert when it breaks.
File or API. Pick whatever your IT team will sign off on, and switch later.
Opyra asks for no changes to your ERP. Pick the path your IT team will accept — and switch later if you want.
Export using our layout and upload it. Good for starting today, without a ticket.
Your system posts the pay period over REST with its own key. Good if you already automate.
ERP-agnostic on principle: the extraction layout is generated from your system's real data dictionary, never from guessed table names — and you run it, inside your own environment.
Security here isn't a badge in the footer — it's what keeps one customer from seeing another's data, and what lets anyone reconstruct any number.
Every sensitive table carries a row-level policy per tenant, enforced even against the table owner. Isolation by construction, not by convention.
Extracts are stored encrypted. So are their file names.
Every query gets a traceable id: who asked, what ran, how long it took, what came back.
New queries, book delivery, and source publication all pass through a person first. Nothing publishes itself.
Operators, managers, and admins see different screens — configurable per customer.
With no approved rate table in force, recalculation declares itself unavailable. With no calibration, a rule declares itself unevaluated. Opyra would rather say “—” than say zero.
Personal data handling is described in the privacy policy (LGPD is Brazil's data protection law). Read the policy →
Schedules — the right report in your inbox, on the agreed date · Targets & alerts — the breach reaches WhatsApp the same day · Briefing — the last-minute meeting, ready in minutes
A recommendation made here <b>comes back with a status in the next edition.</b>
This is why the presentation was rebuilt: six fronts were born after the previous version.
One person's entire working life on a timeline, pulled from the ERP on demand. A life band and a career river: two readings that don't repeat each other.
Seven new rules read straight out of the ERP's own close routine. They catch on the 3rd the setup error that would surface on the 30th of next month.
An X-ray of payroll item by item, using the ERP's canonical calculation id instead of whatever name each company typed. Bases and money never add up together.
17 chapters with dynamic numbering, one shared skin, a timekeeping chapter, a workforce portrait, and the value delivered this month.
A card per metric and a field × metric matrix: which ERP field switches on what, measured by running it rather than guessing.
Accident-risk and social-fund rates checked against the official tables, with the gap in reais and the regulatory sources behind it.
the score opens the document with the ruler attached: it starts at 100 and comes down factor by factor, each with its cap — and the auditor's verdict beside it.
| 2026-03 | +22 / -17 |
|---|---|
| 2026-04 | +19 / -21 |
| 2026-05 | +26 / -15 |
| 2026-06 | +24 / -14 |
| 2026-07 | +30 / -12 |
| 2026-08 | +28 / -11 |
| critica | 0 |
|---|---|
| alta | 1 |
| media | 6 |
| Baby boomers | 34 entram · 31 saem |
|---|---|
| Geração X | 142 entram · 128 saem |
| Millennials | 298 entram · 276 saem |
| Geração Z | 176 entram · 163 saem |
| Money that stayed in | R$ 28.4k |
|---|---|
| Penalty avoided | R$ 7.2k |
| Exposure reduced | R$ 2.4k |
The whole month on one page — <b>and the approver signs the cover.</b>
and the document closes with the appendices: the full profile for reference, and the data governance — where every number came from and who approved it.
the four charts above are the Book's real functions, running on the demonstration company
Upload a payroll extract and Opyra returns your Health Score and aggregate findings on the spot. No personal data, no install, no commitment.
Upload an extract. Results appear on screen, processed in memory and never stored.
We connect one real month and run the whole close alongside your team.
Schedules, targets, the monthly book, and the auditor running on every extract.
Opyra is licensed per seat, under a contract per customer. Because the number depends on your size and which fronts you turn on, we don't publish a table: the quote follows the assessment, with the scope written down.
The same sentence we opened with. Now you know what's behind it.