Skip to content
Hlebourg
Catalogue

Three lines, nine products, one interface

Each line solves a problem our consulting practice kept rebuilding by hand. Each product runs in your own environment and exposes an MCP server, so an agent can operate it without a bespoke integration.

Confidential document control

Nocturne

Encryption, per-recipient tracing and key material for documents that leave your perimeter and must not be untraceable when they do. Built for the files that a DLP product never sees, because they were sent deliberately.

Why it exists
Most leaks are not break-ins. A file is shared legitimately with eleven people, and one of them forwards it. Perimeter tooling stops the first case and is blind to the second.
Who buys it
Legal, corporate finance, deal advisory, IP-heavy engineering

Controlled distribution you own, opened in an ordinary browser

Vault moves from encrypting files on one device to controlling the files you share with other organisations. It is a document room you host yourself, air-gapped if needed. Files render on the server, so readers need only an ordinary browser, and an original is released only where policy permits. Access is decided against attributes, and a refusal names the clause that failed. Every grant carries an expiry. Every open, print attempt and download enters an append-only trail. Your own staff can read offline in an optional desktop viewer.

MCP server
hlebourg-vault
Priced by
One deployment for the organisation, sized on its document estate

Encryption at rest for files on macOS and Windows, with keys protected by the device’s own hardware

Vault is an embeddable component that encrypts files at rest. Each file has its own key, wrapped by a master key, which a root key protects. The root key is held in the Secure Enclave on macOS or as a TPM-backed key on Windows, with the platform’s secure key storage as the fallback. Authenticated AES-256 means a file altered or taken out of its authorised context fails to decrypt with a clear error. The master key rotates every 90 days by default, without downtime.

Platforms
macOS · Windows, offline
You receive
Source code, libraries, optional installers, administrator command-line tool, guides

Per-recipient watermarks that survive a re-save, a print, a photograph and — where enough text remains — a rewrite

Trace watermarks every copy of a PDF, Word, PowerPoint or plain-text document for a registered recipient and records which copy went where and when, through an API you build into your product. Each copy carries its own invisible arrangement in the file, on the rendered page and, where your style rules permit, in the wording, plus a visible tiled overlay. Where enough of a leaked scan, photograph or chatbot summary survives, Trace returns candidate recipients and a confidence figure — never one name as fact. Self-hosted by default.

MCP server
hlebourg-trace
Priced by
The uniquely marked copy, one per recipient per issue

PDF watermarks that tie every copy to a person and a moment, designed to survive printing and re-saving

Trace watermarks a PDF automatically when it is viewed, exported or printed. Marks can be visible or semi-invisible — tiled or diagonal text and a QR code carrying a payload signed with HMAC-SHA-256 — and the same trace can go into XMP metadata. Individual marks carry user ID, email, timestamp, role and profile; generic marks carry only a notice. Either kind, or both, is set per document category, process, application or user group, without code. A verification tool shows whose copy it was and when.

Platforms
macOS · Windows, offline
You receive
Source code, libraries, language bindings, optional installers, guides

One encryption layer for every application you ship, reporting which keystore each device actually provided

Entropy grew out of the Nocturne randomness module into an embeddable encryption SDK and a key service you run yourself, air-gapped if you need it. It encrypts fields and files before they leave the device and wraps data keys with keys you hold. The module refused weaker sources; Entropy does not hide weaker storage. It reports which keystore each device obtained — Secure Enclave or TPM-backed CNG where present — and grades each key hardware-attested, hardware-claimed or software. Its FIPS provider is in the CMVP queue, not yet certified.

MCP server
hlebourg-entropy
Priced by
One production environment, counted as the contract names them; development, staging and CI free — the unit meters production, and rehearsal on your own infrastructure is never a second invoice.

One trusted source of randomness for every application you ship

Entropy gives macOS and Windows applications, and the other Nocturne modules, one interface to cryptographically secure randomness: random bytes of any length, bounded integers whose range and bias are documented, nonces with safe defaults and standards-conformant UUIDs. It draws only on the operating system’s own generator — SecRandomCopyBytes on macOS, BCryptGenRandom on Windows — with no custom generators, no home-made mixing and no fallback to weaker sources. Continuous health tests fail closed and alert the operator without slowing normal calls.

Platforms
macOS · Windows, offline
You receive
Source code, libraries, language bindings, optional installers, guides
Workforce sustainability

Cadence

Burnout screening, team pulse and personal capacity tracking that measure a person against their own baseline instead of a company average. Designed so that the number a manager sees is one the employee has already seen.

Why it exists
Attrition is visible in the resignation letter and in almost nothing that precedes it. Annual engagement surveys report on a quarter that has already been paid for.
Who buys it
People operations, engineering leadership, occupational health

Burnout screening measured against a person’s own baseline, with results the person sees first

Signal scores a screening questionnaire derived from the Copenhagen Burnout Inventory: personal, work-related and client-related exhaustion on a 0 to 100 scale, in a green, amber or red band. It reads each result against that person’s own 90-day baseline, not a company average. Short check-ins every three to four weeks and work-pattern metadata from your AI gateway feed a drift index; message content is never read, emotion never inferred. The person sees their own file first; a manager sees a band only with consent. Self-hosted by default. Screening, not diagnosis.

MCP server
hlebourg-signal
Priced by
Each person in scope of screening, counted once a month

Burnout screening that stays on the employee’s device

Signal is an embeddable burnout-screening module. On the employee’s device it runs a questionnaire built on the structure of the Copenhagen Burnout Inventory. It scores personal, work-related and client-related exhaustion, combines them into a burnout index from 0 to 100 and places it in a green, amber or red band with non-clinical suggestions. Trends cover rolling periods you choose, such as 30, 90 or 180 days. Thresholds, suggestions, reminders and questionnaires are yours to configure; each person can adjust or switch off reminders.

Platforms
macOS · Windows, offline
You receive
Source code, sample configurations, tests, guides; screens not included

Short check-ins for a whole workforce, measured on what changed, not on what was asked

Pulse takes the short check-in from one person’s device to a whole workforce: a round every four to twelve weeks, never faster than you can act on it, by email, Slack, Teams, SMS, WhatsApp, QR or kiosk. Answers become an engagement index from 0 to 100 with a driver breakdown. Locked core wording lets each team be read against its own history, and a flag marks jumps that history does not explain. HR can answer a comment without learning who wrote it. Self-hosted, the data stays on your servers.

MCP server
hlebourg-pulse
Priced by
The organisation's employee perimeter, counted once

Cadence Pulse

Available

Short motivation check-ins, scored on the device

Pulse is a motivation check-in you build into your own application, scored and kept on the employee’s device. A weekly check-in has three to five questions and a monthly one has ten to twelve, in versions for different groups or topics. It scores reverse-worded questions correctly and applies the weights you set. The answers become a Motivation Index from 0 to 100, with trends over the last 30 and 90 days. Hints come from rules you write, and each person can switch notifications off.

Platforms
macOS · Windows, offline
You receive
Source code, sample configurations, tests, guides; screens not included

AI spend attribution that reports what it could not attribute and never ranks the people behind the spend

Ledger joins gateway spend logs, chat admin APIs and IDE usage reports to your org graph, so AI cost lands on a named cost centre and stays there through key rotation and leavers. It traces agent fan-out to the person who started the chain and counts directed and reviewed agent calls per week at team level by default, never as a ranking. Every total shows its unattributed share, and why. Self-hosted by default; it reads metadata only, never prompts or completions, and one call erases a person’s data.

MCP server
hlebourg-ledger
Priced by
One attribution perimeter, counted in people whose agents generate the spend

A personal effectiveness index built from what people report, not from what they are tracked doing

Ledger combines a daily self-rating of energy, focus and completed tasks with the person’s Pulse Motivation Index and the inverse of their Signal burnout index, where available. It smooths the result over a window such as seven or fourteen days and returns an Effectiveness Level Index from 0 to 100, a plain level such as satisfactory, at risk or critical, and a daily card with one suggestion. Rating scales are configurable; weights and window adjust within agreed limits. One person’s data can be deleted.

Platforms
macOS · Windows, offline
You receive
Source code, sample configurations, tests, guides; screens not included
Offline-first operations

Harbour

A counterparty registry, an order and demand engine, and an obligation follow-up system that keep working with the network cable pulled out. State lives on the machine; synchronisation is something you choose, not something you depend on.

Why it exists
Operations software assumes constant connectivity and a head office. Warehouses, sites, clinics and vessels do not have either, and the spreadsheet that fills the gap is never reconciled.
Who buys it
Operations, trade and distribution, field and site services

An encrypted offline record base that synchronises across devices, with every merge rule declared per field

Registry is an encrypted record base for work orders, inspections, stock counts and approvals; it works without a network and synchronises across devices and with your own backend. Each field is declared in a versioned entity model with its own merge rule, such as a state machine for statuses. Where no rule describes a conflict, Registry quarantines the record rather than guessing. Every mutation is audited with its origin device, so a field’s current value can be traced to the competing writes and the rule applied. Self-hosted by default.

MCP server
hlebourg-registry
Priced by
The autonomous site — one place, vessel or enclave that has to keep working through a disconnection, counted where it is, not by what is in it

An encrypted client base that works without a network

Registry is a client base built into your own macOS and Windows applications. It keeps encrypted client cards — contact channels, company details, the responsible person and team, tags, categories, statuses, last contact, next reminder and attached files — on the user’s device. Fields and lists are configurable. Search covers the fields you choose, with filters by tag, category, status and date. Every change leaves an audit entry: who, when and which fields. Selected clients export to CSV or JSON, and reports to PDF.

Platforms
macOS · Windows, offline
You receive
Source code, configurations, tests, guides; compiled components where agreed

Order history, quantities and confirmed dates, offline — conflicts settled by a state machine, not a merge

Flow keeps a client’s orders as commitments — quantity, place and date — each in one declared state. Every change is a transition with actor and time. Devices work offline and synchronise to a backend inside your boundary, under one state machine. Concurrent changes get a legal interleaving or a named rejection, never a record in two states. Flow reports how often a client’s confirmed dates held (descriptive, not a forecast) and exports the commitment log for audit. Not an ERP: it does not invoice or post to a ledger.

MCP server
hlebourg-flow
Priced by
One field operation, counted in operators who commit offline

Harbour Flow

Available

Order history, volumes and repeat-order dates, offline

Flow records each client’s orders — lines, quantities, prices, discounts, taxes, configurable statuses and internal comments — in an encrypted local database. It totals quantity and value by month, quarter or year, shows average order size and flags unusual rises and falls under rules you set. From the interval between past orders it proposes a repeat-order date per client and item, so a reminder can go out before the date. Every calculation runs on the device, and Flow never sends client data to outside servers.

Platforms
macOS · Windows, offline
You receive
Source code, configurations, tests, guides; compiled components where agreed

Every open obligation, its owner and its deadline, with escalation that works offline and a count of obligations nobody owns

Watch is the ledger of what is owed, by whom, by when, and who currently owns it — renewal windows, regulatory filings, inspection intervals, contractual cure periods — loaded by you or fed by what Flow signs off. Reminders become notices that escalate by rules running on the device, so a site with no connectivity still escalates on time. Snooze is capped and attributed, recurrence follows the schedule, and notice lead time is measured against policy. The headline figure counts obligations whose owner has left your directory or changed role.

MCP server
hlebourg-watch
Priced by
One register under management — the whole obligation estate of one organisation, at whatever size and whatever date it is loaded

Harbour Watch

Available

Every call, meeting and open request from a client, with reminders that work offline

Watch logs calls, meetings, emails and notes against a client and, where it matters, an order, with interaction types and their fields defined in configuration. Needs and tasks carry a priority, a status, a due date and a responsible person. Reminders are calculated on the device from due dates, repeat-order dates and your rules — with lead times, snooze and recurrence — and handed to the operating system’s notifications. Contacts can be imported from the device’s address book, with the user’s permission.

Platforms
macOS · Windows, offline
You receive
Source code, configurations, tests, guides; compiled components where agreed

Not sure which line you need

Describe the failure you are trying to prevent and we will tell you which product applies — including when the answer is none of them.